UnThreaded | Threaded | Whole Thread (135) | Ignore Thread Prev Thread | Next Thread
Author: dividends20 Two stars, 250 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: of 253091  
Subject: Does this stuff work ? Date: 1/28/2013 10:44 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Hi: I am new to mechanical investing. Seems pretty easy. Identify a screen and repeat.

Q - Does this stuff work in real life. Has anyone invested using this method and beaten S&P 500 index over a long period ?
Print the post Back To Top
Author: musselmant Three stars, 500 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241337 of 253091
Subject: Re: Does this stuff work ? Date: 1/28/2013 11:00 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 38
1. no
2. not very well
3. not as well as people on here say
4. not as well as the backests say
5. the backtests are flawed
6. the data the backtests is based on are flawed
7. results are more random than you might think
8. reaults don't take into account black swans
9. let alone the worst 12 year period in 150 years of stock market history
10. results don't include all start dates, actual transaction costs over time, taxes, inflation, mistakes in placing orders, broker errors, political and legal developments that change the basis for what works and why
11. we hope they work anyway, because we are greedy and aren't happy with index fund returns, and we are self-deluded

Print the post Back To Top
Author: FLARAM Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241338 of 253091
Subject: Re: Does this stuff work ? Date: 1/28/2013 11:31 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 6
This question comes up repeatable as web surfers invariably drop by and notice what appears to be an easy way to beat the market. But as musselmant suggested above it's neither as good nor as easy as it looks. My reply to essentially the same question by another person last May hasn't changed. See

Backtest vs reality
http://boards.fool.com/for-those-of-you-who-have-successfull...

I wouldn't agree with musselmant and say it defiantly doesn't work as I am still a few points ahead of the game but judging by the large number of people who have come and gone from the board I have to consider people like myself may have just had a lucky roll of the dice.

I am in the camp of one that believes numbers aren't quite as manipulated as the perfumed pigs corporate management and the financial advice industry is capable of presenting.

RAM

Print the post Back To Top
Author: LonghornBoy Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241339 of 253091
Subject: Re: Does this stuff work ? Date: 1/28/2013 11:37 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Seems pretty easy. Identify a screen and repeat.

Bwahahahahahahahahahahahahahahahahahaha!!!

<pauses for a breath...>

Hahahahahahahahahahahahahahahahahahahah!!!!

<loses control and falls on floor...>

Ah...hahahahahahahahahaha...

<wipes tears from eyes...>

Thanks...that was the best laugh I've had in a while.

As I'm sure many here will attest, you will second guess your screen to no end when the market zigs while your screen zags. You'll be overcome with the urge to tinker, pull money out, swap screens, and do all sorts of stuff to try and make it better.

Easy? Not on your life... ;)

Print the post Back To Top
Author: Huskerfan101 Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241341 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 12:42 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
Here is my two cents worth. I have been lurking here for at leat a couple years and have learned a lot from the other folks here. One, is it easy, no, but it's probably easier now then in the past. Over the last 10-12 years a lot of painful lessons have been learned, namely never, never, never, underestimate risk and your ability to handle it, a lesson recently relearned by me (thanks FB). The methods outlined here to avoid bear or minimize bear markets is a key development in the last 4 years. Also since the screens are so much older many have meaningful out of sample periods, so you can really get an understanding of real world performance. With all that said here are my results since I started earnestly doing it, made much easier by FoliFn

My start date is 04/09/2012, results from there till the end of the year are 17.07% VS S&P 4.92(according to FolioFNn). Through COB today i'm up 24.11 vs S&P 10.67, again numbers as reported in my account.

So is it "easy" for me, so far yes. After a market correction or period of underperformance, we'll see. I owe all this to the members of this board, so if you pursue MI and succedd remember these guys were the trail blazers and made it much easier for those who follow.

To the members of the board, i do want to add a SIPRO screen, can you recommend one.I'm considering small value or sliver parachutes, are they suitable for FolioFn window trades, in that I wont get killed with the spreads?

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: HankDfrmSD Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241342 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 2:37 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Check these poll results for 2012

http://boards.fool.com/poll-2012-mi-returns-30460058.aspx?so...

Print the post Back To Top
Author: joelxwil Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241344 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 9:58 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Well, yes and no. My main objection to the MI scheme is that it involves fixed holding periods. There is absolutely no way to tell how long you should hold a stock when you buy it. You need to be able to react to situations as they unfold. That is why the MI systems that I have traded all have stops and do not have minimum holding periods. But I use Fasttrack to build mechanical systems, and that allows for stops, etc.

Print the post Back To Top
Author: Volition One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241347 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 4:12 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
My main objection to the MI scheme is that it involves fixed holding periods. There is absolutely no way to tell how long you should hold a stock when you buy it. You need to be able to react to situations as they unfold. That is why the MI systems that I have traded all have stops and do not have minimum holding periods.

Have you backtested your method? Or do you use subjective factors to decide when to sell?

Print the post Back To Top
Author: joelxwil Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241348 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 4:43 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Fastbreak is a backtesting program. You can rank by a number of methods, buy however many stocks you want to hold, and you can put in stops. You can also input market timing algorithms. I use Trade to generate market timing systems, which can then be input into Fasttrack. Both Fastbreak and Trade work on the Fasttrack database of stocks.

However, the PRISM method, http://www.actwin.com/kalostrader/VisualMethods.htm, is a very simple method of market timing also. It is simply a trend following method.

Print the post Back To Top
Author: JLC Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241351 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 6:15 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Q - Does this stuff work in real life. Has anyone invested using this method and beaten S&P 500 index over a long period ?

A - it depends on what your goals are. If its to turn $1000 into $1M in a couple of years and retire at 25, then no. If its to beat the S&P by 1-2%/year, possible. If its to help decrease risk and variability while getting S&P returns, then yes.

JLC

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241353 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 6:51 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 31
Q - Does this stuff work in real life. Has anyone invested using this method and beaten S&P 500 index over a long period ?

I think it works. It takes discipline more than anything else, and a judicious choice of screens.

I've had an annualized return of 15% over the last 15 years. 15% doubles your money roughly every five years. The S&P index has returned 4.5% annualized over the same period. So it doubled once in 15 years.

Inflation over the same period was 43%. Which means the real total return of the S&P was about 40%. My MI real total return was about 460%.

Elan

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241357 of 253091
Subject: Re: Does this stuff work ? Date: 1/29/2013 10:48 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 7
My main objection to the MI scheme is that it involves fixed holding periods.
No they don't. What the MI screens do is have fixed re-evaluate (think: rebalance) periods.

That is why the MI systems that I have traded all have stops and do not have minimum holding periods.
The MI holding period is generally "as long as it continues to rank in the top N positions."

As far as stops --- amateurs & beginners tend to think that this is a successful technique. It isn't.

A couple of quotes on this point:
"Using stops is the definition of risk management for some. Using stops is 10th grade risk management." -- Peter Brandt

"In markets, we are paid for assuming the correct risks at the correct time, and managing them appropriately. Many times, understanding what those correct risks are requires some counter-intuitive thinking, and things that might seem to be prudent risk management (e.g., very tight stops) often amount to virtually certain losses over a large sample size." -- adamhgrimes

Print the post Back To Top
Author: ges Three stars, 500 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241361 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 10:41 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
As far as stops --- amateurs & beginners tend to think that this is a successful technique. It isn't.

Based on a lot of testing, I agree.

Print the post Back To Top
Author: culcha Big gold star, 5000 posts Old School Fool CAPS All Star Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241362 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 10:45 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 3
Seems pretty easy. Identify a screen and repeat.

Pretty easy in some ways. Like losing weight. Plan a diet of small portions of nutritious food, get regular exercise, and repeat.

What's the problem???

Anybody can do it.

culcha

Print the post Back To Top
Author: mungofitch Big gold star, 5000 posts Top Favorite Fools Top Recommended Fools Feste Award Winner! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241363 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 11:02 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 33
Well, yes and no. My main objection to the MI scheme is that it
involves fixed holding periods. There is absolutely no way to tell how
long you should hold a stock when you buy it.


For whatever it's worth, I consider this by far the best strength of MI.
Why?
Nobody has found another method of deciding when to sell that works well
on average, or in backtest, or in real live, or even with decent anecdotes.
A fixed hold period rules out all the other reasons to sell a stock at
a given time, virtually all of which either fail on average when done
repeatedly because they're based on emotion or poor pseudo-theories,
or can't even be tested as a thought experiment because they are arbitrary.
It also ensures that the selling date criteria are quantifiable and testable.
If you're going to do something, do something that works on average in backtest.
It might not work in real life, but it makes NO sense to do something
that didn't work on average in the past, or something that can't even
be tested hypothetically on how it might have done in the past.
I take fairly broad definitions of those tests, but hold to them rigorously.

Whenever you buy something, you should write down why you bought it,
what will trigger you to sell it, and what good evidence makes you think
that that pair of rules will make money on average through time.
If you can't answer those three questions, don't buy a stock.

Jim

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: joelxwil Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241364 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 11:32 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Nobody has found another method of deciding when to sell that works well on average, or in backtest, or in real live, or even with decent anecdotes.

Not true. Whenever I generate a system, it is improved by putting in stops. It improves both the overall result and lessens the drawdowns. You are just not doing it right if your stops do not improve things.

Print the post Back To Top
Author: JeanDavid Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241365 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 12:31 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Nobody has found another method of deciding when to sell that works well on average, or in backtest, or in real live, or even with decent anecdotes.

Not true. Whenever I generate a system, it is improved by putting in stops. It improves both the overall result and lessens the drawdowns. You are just not doing it right if your stops do not improve things.


Hmmm. Yet Jim lives in Monaco and Menton, and you live in Maryland.

Print the post Back To Top
Author: klouche Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241366 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 3:21 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Hmmm. Yet Jim lives in Monaco and Menton, and you live in Maryland.

You mean the Commonwealth of Maryland?

I've tested stops in the past, starting back from when Peter Kuperman did so.

Seems like the advantage, if any was too slight to interest in implementing. On the stop loss side, ~18% may give an advantage for the right screen and volatility profile. On the profit latch side, it was more promising and I use them on my index holdings, however in the low single digits.

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241371 of 253091
Subject: Re: Does this stuff work ? Date: 1/30/2013 10:19 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
As far as stops --- amateurs & beginners tend to think that this is a successful technique. It isn't.

Based on a lot of testing, I agree.


Unhappily, I learned it in one lesson.

OXHP, Oxford Health Plan, a TMF pick for a "buy and hold forever" stock.

It rose to 60, and I thought I'd protect my profit by selling a stop-loss order at 50. It bobbled around 60 for a while. Then the auditors announced "accounting irregularities". Translation: The books are cooked; the claimed income is all lies."

OXHP opened the next morning at 30 and went down from there. IIRC, my stop-loss limit order got filled at 20. An hour later the stock was back up to about 30.

Print the post Back To Top
Author: mungofitch Big gold star, 5000 posts Top Favorite Fools Top Recommended Fools Feste Award Winner! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241392 of 253091
Subject: Re: Does this stuff work ? Date: 2/1/2013 11:30 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 12
As far as stops --- amateurs & beginners tend to think that this is a successful technique. It isn't.
...
Based on a lot of testing, I agree.
...
Unhappily, I learned it in one lesson.


There are lost of different investing strategies in the universe.
Whether you're Warren Buffett or somebody listening to (ugh) Cramer, there is a trigger for a purchase and a trigger for a sale.
The trigger might be simple or complex, and may or may not involve a stop loss.
Each entire combination of rules has to be considered as a set when discussing whether something "works" or not.

As far as I know adding stop losses does not improve results on any quant screening
approaches similar to those discussed here, based on a lot of research.
Stops may add value when you're using hand selected stocks by some
"expert" on the internet, or in other situations, I don't know.
Different questions, different answers.

Based on the evidence, for what we do here, if you use stops you'll do worse.
If you don't believe the output of testing, don't use this family of approaches at all.

Jim

Print the post Back To Top
Author: DoesMIWork One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241393 of 253091
Subject: Re: Does this stuff work ? Date: 2/2/2013 8:03 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Agree. The only stop I have is during the rebalance- once a week, once a month, etc. That's why it's all mechanical. You sleep better at niight.

Actually, the only exception is if your pick falsified their numbers, become under criminal investigation, etc.

Print the post Back To Top
Author: warrl Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241400 of 253091
Subject: Re: Does this stuff work ? Date: 2/2/2013 4:53 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Actually, the only exception is if your pick falsified their numbers, become under criminal investigation, etc.

And that's why my dividend growth screen has a couple of technical steps dropped into the midst of all its fundamental steps.

Enron didn't pass that screen before the fraud was discovered, and didn't qualify for it after the *real* fundamentals came out. But when the reported fundamentals were known to be the result of fraud, it appeared to be the top pick with well over 100% trailing-twelve-month dividend yield. Until I added a "the market seriously disrespects this stock, and there's probably a good reason" step or two.

I'd recommend some similar filters in any other fundamental screen.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241546 of 253091
Subject: Re: Does this stuff work ? Date: 2/12/2013 6:15 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
mungofitch (241363)

There is absolutely no way to tell how long you should hold a stock when you buy it.

My Sell-to-Buy system might be relevant to this thread, as it deals with the selling dilemma:

It works well for long-term investment (validated by backtesting 5 year time periods over a 10 year daily dataset on the UK FTSE 100 constituent companies) when buys are made at an average of about one every two weeks. So in the short term early results may lose out - but the system appears to work in the long term long term probably due to the cyclical nature of stock prices. All profits and dividends should be accumulated and no withdrawals being made so that additional stocks may be acquired. I have tested it with 3, 5 and 10 starting holdings and although it might intuitively seem higher risk, the best result is with just 3 initial holdings. It could be useful for investment clubs interested in long term growth.

1. Divide an initial investment sum by 3 to give a lot size for all subsequent purchases.
2. Buy your first three holdings.
3. To buy your 4th holding, only do so if you best holding is showing at least 5% profit.
4. Bank the profit and continue to accumulate profits each time you sell.
5. Buy next holding from banked profits when possible
6. If bank is insufficient and no holding is in profit when you want to buy, do nothing.
7. Apply a 40% stoploss to each holding.
8. Sell all on market crash (10/10/2008 and 3/3/2009).
9. Track your stoploss companies (including those you did not buy) and rebuy if they fall below 50% of original price.


It works for me and if you have a record of your transactions over the last 5 years you might like to try the scheme to see how it could have performed for you.

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241560 of 253091
Subject: Re: Does this stuff work ? Date: 2/12/2013 5:36 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
My Sell-to-Buy system might be relevant to this thread, as it deals with the selling dilemma:

This set of rules is almost completely hindsight bias and Look-Ahead Bias, and hence useless.

8. Sell all on market crash (10/10/2008 and 3/3/2009).
Define "market crash" -- in such as way that it can be identified at the time, and not just in hindsight. Do you mean to sell _before_ the crash? Good luck being able to predict the near-term future.
Or do you mean to sell _after_ the crash? Which means that you lock in your loss and miss out on the gain if you've sold at the bottom.

So after you've sold, how do you know when to buy again? These rules don't say anything about that.

To buy your 4th holding, only do so if you best holding is showing at least 5% profit.
Huh? What does stock #1 (hest holding) have to do with stock #4?
Where does the money come from to buy stock #4?

Track your stoploss companies (including those you did not buy) and rebuy if they fall below 50% of original price.
What does "those you did not buy" mean? Is this (these?) the #4(s) that you didn't buy because nothing went up 5% (#3)?
And why, oh why would you buy a stock that lost 50%???? Some sort of belief that what goes down must come up? 'cause that's not true at all.

How long do you keep track of the #4 stocks that you haven't bought yet? Do you periodically re-evaluate? If not, why not? It doesn't make sense to buy a stock _now_ just because you liked it a year ago. IF you periodically re-evaluate the desirability of the (unbought) #4, why wouldn't you also periodically re-evaluate the other 3 stocks that you did buy?

These rules only say to sell if a stock loses 40%, or if the market crashes. That seems rather incomplete. What if a stock goes nowhere, or goes down 10%=20% and stays there?

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241567 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 6:56 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
Rayvt (241560)

You have raised lots of points and I will do my best to answer them.

My Sell-to-Buy system might be relevant to this thread, as it deals with the selling dilemma:

This set of rules is almost completely hindsight bias and Look-Ahead Bias, and hence useless.


The rules were developed in hindsight 5 years ago on the basis of analysing the previous 5 years of data.

Since then they have continued refinement and I am currently using them with great effect on four real

portfolios with great results.


8. Sell all on market crash (10/10/2008 and 3/3/2009).
Define "market crash" -- in such as way that it can be identified at the time, and not just in hindsight. Do

you mean to sell _before_ the crash? Good luck being able to predict the near-term future.
Or do you mean to sell _after_ the crash? Which means that you lock in your loss and miss out on the gain if

you've sold at the bottom.


I identify a market crash by tracking the trend of the index value. I declare a crash when the trend of the

previous 20 days reaches the -0.5% per day mark. You then start buying again from the bottom of the market.

So after you've sold, how do you know when to buy again? These rules don't say anything about that.

Just carry on with whatever stock picking system you use, but apply the rules I have given about whether to buy and what to sell.

To buy your 4th holding, only do so if you best holding is showing at least 5% profit.
Huh? What does stock #1 (hest holding) have to do with stock #4?
Where does the money come from to buy stock #4?


There is no connection between the stocks. The purpose is to take profit when available at the time you

want to buy another, so stock#4 is bought from the proceeds of selling the best of the three holdings (all

stock purchases are made to the same purchase value). Profits are retained and grow so that you can buy an

additional holding. I continue to update the dataset for my 5-year-to-date model each day and the current

results now have 11 holdings from the original 3 stating holdings.


Track your stoploss companies (including those you did not buy) and rebuy if they fall below 50% of

original price.
What does "those you did not buy" mean? Is this (these?) the #4(s) that you didn't buy because nothing went

up 5% (#3)?


Yes, keep track of ALL those that fall to 40% loss (those you had bought and those you could not buy either

because you had no holdings at >5% profit or because you had not accumulated sufficient in bank.)

And why, oh why would you buy a stock that lost 50%???? Some sort of belief that what goes down must come

up? 'cause that's not true at all.


Are you sure they don't recover? Except for a few disasters that has not been my experience and, contrary

to popular belief, modelling has demonstrated the very significant benefit of rebuy on 50% loss. The 5year

model currently shows 7 holdings without the rebuy, but 11 with rebuy. All the rebuys made a profit of between 16.4% and 47.3%.

How long do you keep track of the #4 stocks that you haven't bought yet? Do you periodically re-evaluate?

If not, why not? It doesn't make sense to buy a stock _now_ just because you liked it a year ago. IF you

periodically re-evaluate the desirability of the (unbought) #4, why wouldn't you also periodically re-

evaluate the other 3 stocks that you did buy?


I have tried to develop a simple system which avoids the difficult issue of re-evaluation. That is a

fundamental part of the system. It may not make sense but it works.

These rules only say to sell if a stock loses 40%, or if the market crashes. That seems rather

incomplete. What if a stock goes nowhere, or goes down 10%=20% and stays there?


If a stock goes nowhere it simply remains a current holding until such time as it meets the sell criteria.

This means that some stocks are held for a long time. The longest held in the 5-year model, Kazakhmys,

lasted 707 days. The shortest was 4 days and the median 111 days (a total of 71 were bought).

Interestingly, Kazakhmys is a case in point as it was sold on 26 September 2011 as a 40% stoploss, but later

created a rebuy on 50% loss on 28 August 2012 which was sold at 18.2% profit after 22 days on 18 September

2012.
It is also worth pointing out that as a long-term stock, Kazakhmys added significant dividends to the bank.
Incidentally, I only apply the rebuy once - if the rebuy hits stoploss cash it in but do not rebuy again.


I hope this makes the process clear enough for you to try it.


Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241569 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 8:58 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
I hope this makes the process clear enough for you to try it.

I'll stand by my remark about look-ahead bias.

Your comment about Kazakhmys was interesting & puzzling. First, yahoo finance doesn't show it on any US market, just London, Hong Kong, etc. So such a stock wouldn't be investable by 99% of us. I think I found it, though, from your description of the price action. That chart is nothing that I would touch.

Second, you say you sold it at 18% profit in 22 days. But I don't see anything in your posted rules that says to sell it. Looks to me that you just do things ad-hoc.

You then start buying again from the bottom of the market.
Sure. That's easy to say. Every beginning investor knows that you should sell at the top and buy at the bottom, so what's hard about making a ton of money?
The only way you know that you are at the bottom is AFTERWARDS, after the market has gone up from that bottom. But when you are there, you don't know if the market goes down or up from there. That's why these beginning investors who thought it was so easy lose their shirt.


Are you sure they don't recover? Except for a few disasters that has not been my experience and, contrary to popular belief, modelling has demonstrated the very significant benefit of rebuy on 50% loss.
This is astonishing to me. All I can gather from this is that your experience must be extremely limited.

Yeah, I'm pretty sure that in general they don't recover. Just thinking about it: In general, a stock doesn't lose 50% for no reason. The company has to be doing really bad. Given the fact that the company did so bad, what is the chance that they suddenly turn around and double? Sure, it occasionally happens. But in general, losers continue to lose and winners continue to win. That's the case for sports teams, countries, and businesses & stocks.

Now, just thinking that something is so doesn't mean that it is so. You need to look at actual data, since reality trumps theory.

So I went to my favorite backtester, stockscreen123, and created a "price crash" screen. My standard filters for volume, price (no penny stocks) and liquidity, then screened for 52 week price loss of 50%. Holding period of 4 weeks, Jan-1999 thru Feb-2013. Maximum of 50 stocks, ranked by yield.

Results:
Annualized Return: -12%
Sharpe Ratio: -0.5
Total return: -85%

So, yeah, I'm pretty sure.

Looking deeper into the backtest results, I can see how you could fool yourself into thinking this is a good strategy.
If you look at a good period, you get some eye-popping returns. 195% 125% 60% 40% These are all 4-week returns! Yummy!!

But most of the period you get returns like -15%, -35%, -84%.

I've seen this movie before, and I know how it ends.
You come up with a great theory, look only at data that supports that theory, look only at a time period where that theory works --- and subsequently get crushed by reality. Oh yeah ... and poo-poo people who try to warn you that you are headed for a cliff.

Been there, done that.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241572 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 11:19 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 14
Rayvt (241569)

I'll stand by my remark about look-ahead bias.

None of my process looks ahead.

Your comment about Kazakhmys was interesting & puzzling. First, yahoo finance doesn't show it on any US market, just London, Hong Kong, etc

My opening post (241546) explained that my dataset uses the UK FTSE 100 constituents. Those are the top 100 companies in UK listed by market capitalisation. I use those companies because they are highly liquid with very small buy/sell ratios.

you say you sold it at 18% profit in 22 days. But I don't see anything in your posted rules that says to sell it. Looks to me that you just do things ad-hoc.


See Rule 3 - it is about when to sell.
It says to sell the best to buy the next. I sold Kazakhmys because it was the best at the time I bought another stock, RBS, which is still a current holding at 33% profit. However, that is beside the point which is to illustrate that losers can become winners.

contrary to popular belief, modelling has demonstrated the very significant benefit of rebuy on 50% loss.
This is astonishing to me. All I can gather from this is that your experience must be extremely limited......You need to look at actual data, since reality trumps theory.


I have been investing in stocks and shares for over 20 years and had just two failures (GEC and Northern Rock). My 5-year model is based on 86 sequential dataruns over the last 10 years covering both rising and falling markets and it is not theory - I have been using it for real for the last 5 years.


You come up with a great theory, look only at data that supports that theory, look only at a time period where that theory works

Not so. All my data is real and not selected in any way.

Enough of all this negativity. I have offered you a system in response to a request at post 214366 (dividends20) that works exceedingly well for me in UK and might conceivably work well for you too. I have nothing to gain - and you might.

The acid test is for you to create a spread sheet and try the scheme on you own buying data from the last 5 years.

I would suggest that anyone involved in, say, an investment club meeting once a month to make a buying decision could do well to try the system on their past data to see if it would have given a better result.

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: JeanDavid Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241574 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 12:29 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 3
I can see how you could fool yourself into thinking this is a good strategy.

Back when I had a full-commission broker, she suggested trading S&P500 Index futures contracts. I put in $7,000 to try it out. All I was willing to lose at the time. She made several trades a day. It was a great strategy. I doubled my money every week for three straight weeks! Wow! But then my investment dropped 50% every week for five straight weeks. She then requested more money because the brokerage firm had a minimum amount for trading those contracts, and I had gotten below it. It did not seem like a good idea, so I put no more money in that. She traded some more on her own personal account and lost a lot more. She really believed in that stuff, but learned slower than I did.

Print the post Back To Top
Author: dackle7 One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241577 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 4:42 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
RayVT wrote: You come up with a great theory, look only at data that supports that theory, look only at a time period where that theory works --- and subsequently get crushed by reality. Oh yeah ... and poo-poo people who try to warn you that you are headed for a cliff.

Been there, done that.


The same could be said for following Faber's 10-month moving average strategy, just because it worked well over the last 10-15 years. Consider the results for an investor following the 10-month SMA strategy from 1982 to 1999. Buy-and-hold would've turned $10,709 into $136,293 plus dividends, whereas the Faber follower would've ended with only $51,703 plus dividends, but less taxes and commissions.


S&P's S&P/Faber
Date S&P 10m SMA Faber Ratio
Jul 1982 107.09 116.13 107.09 1.000
Aug 1982 119.51 115.89 107.09 .896
Sep 1982 120.42 115.30 107.91 .896
Oct 1982 133.71 116.41 119.81 .896
Nov 1982 138.54 118.23 124.14 .896
Dec 1982 140.64 120.98 126.02 .896
Jan 1983 145.30 124.31 130.20 .896
Feb 1983 148.06 127.48 132.67 .896
Mar 1983 152.96 131.58 137.06 .896
Apr 1983 164.42 137.07 147.33 .896
May 1983 162.39 142.60 145.51 .896
Jun 1983 168.11 147.46 150.64 .896
Jul 1983 162.56 151.67 145.67 .896
Aug 1983 164.04 154.70 146.99 .896
Sep 1983 166.07 157.46 148.81 .896
Oct 1983 163.55 159.75 146.55 .896
Nov 1983 166.40 161.86 149.11 .896
Dec 1983 164.93 163.54 147.79 .896
Jan 1984 163.41 164.59 146.43 .896
Feb 1984 157.06 163.85 146.43 .932
Mar 1984 159.18 163.53 146.43 .920
Apr 1984 160.05 162.73 146.43 .915
May 1984 150.55 161.52 146.43 .973
Jun 1984 153.18 160.44 146.43 .956
Jul 1984 150.66 158.90 146.43 .972
Aug 1984 166.68 159.21 146.43 .878
Sep 1984 166.10 159.18 145.92 .878
Oct 1984 166.09 159.30 145.91 .878
Nov 1984 163.58 159.31 143.70 .878
Dec 1984 167.24 160.33 146.92 .878
Jan 1985 179.63 162.38 157.80 .878
Feb 1985 181.18 164.49 159.17 .878
Mar 1985 180.66 167.50 158.71 .878
Apr 1985 179.83 170.17 157.98 .878
May 1985 189.55 174.05 166.52 .878
Jun 1985 191.85 176.57 168.54 .878
Jul 1985 190.92 179.05 167.72 .878
Aug 1985 188.63 181.31 165.71 .878
Sep 1985 182.08 183.16 159.96 .878
Oct 1985 189.82 185.42 159.96 .843
Nov 1985 202.17 187.67 170.36 .843
Dec 1985 211.28 190.68 178.04 .843
Jan 1986 211.78 193.79 178.46 .843
Feb 1986 226.92 198.50 191.22 .843
Mar 1986 238.90 203.44 201.32 .843
Apr 1986 235.52 207.80 198.47 .843
May 1986 247.35 213.45 208.44 .843
Jun 1986 250.84 219.67 211.38 .843
Jul 1986 236.12 225.07 198.97 .843
Aug 1986 252.93 231.38 213.14 .843
Sep 1986 231.32 234.30 194.93 .843
Oct 1986 243.98 237.57 194.93 .799
Nov 1986 249.22 241.31 199.11 .799
Dec 1986 242.17 242.84 193.48 .799
Jan 1987 274.08 246.35 193.48 .706
Feb 1987 284.20 251.22 200.63 .706
Mar 1987 291.70 255.66 205.92 .706
Apr 1987 288.36 259.41 203.56 .706
May 1987 290.10 264.81 204.79 .706
Jun 1987 304.00 269.91 214.60 .706
Jul 1987 318.66 278.65 224.95 .706
Aug 1987 329.80 287.23 232.82 .706
Sep 1987 321.83 294.49 227.19 .706
Oct 1987 251.79 295.45 177.75 .706
Nov 1987 230.30 291.07 177.75 .772
Dec 1987 247.08 287.36 177.75 .719
Jan 1988 257.07 283.90 177.75 .691
Feb 1988 267.82 281.85 177.75 .664
Mar 1988 258.89 278.72 177.75 .687
Apr 1988 261.33 274.46 177.75 .680
May 1988 262.16 268.81 177.75 .678
Jun 1988 273.50 263.18 177.75 .650
Jul 1988 272.02 258.20 176.78 .650
Aug 1988 261.52 259.17 169.96 .650
Sep 1988 271.91 263.33 176.71 .650
Oct 1988 278.97 266.52 181.30 .650
Nov 1988 273.70 268.18 177.88 .650
Dec 1988 277.72 269.17 180.49 .650
Jan 1989 297.47 273.03 193.32 .650
Feb 1989 288.86 275.78 187.73 .650
Mar 1989 294.87 279.05 191.63 .650
Apr 1989 309.64 282.67 201.23 .650
May 1989 320.52 287.52 208.30 .650
Jun 1989 317.98 293.16 206.65 .650
Jul 1989 346.08 300.58 224.92 .650
Aug 1989 351.45 307.83 228.41 .650
Sep 1989 349.15 315.37 226.91 .650
Oct 1989 340.36 321.64 221.20 .650
Nov 1989 345.99 326.49 224.86 .650
Dec 1989 353.40 332.94 229.67 .650
Jan 1990 329.08 336.37 213.87 .650
Feb 1990 331.89 338.59 213.87 .644
Mar 1990 339.94 340.53 213.87 .629
Apr 1990 330.80 341.81 213.87 .647
May 1990 361.23 343.33 213.87 .592
Jun 1990 358.02 343.99 211.97 .592
Jul 1990 356.15 344.69 210.86 .592
Aug 1990 322.56 342.91 190.97 .592
Sep 1990 306.05 338.91 190.97 .624
Oct 1990 304.00 333.97 190.97 .628
Nov 1990 322.22 333.29 190.97 .593
Dec 1990 330.22 333.12 190.97 .578
Jan 1991 343.93 333.52 190.97 .555
Feb 1991 367.07 337.15 203.82 .555
Mar 1991 375.22 338.54 208.35 .555
Apr 1991 375.35 340.28 208.42 .555
May 1991 389.83 343.65 216.46 .555
Jun 1991 371.16 348.51 206.09 .555
Jul 1991 387.81 356.68 215.34 .555
Aug 1991 395.43 365.82 219.57 .555
Sep 1991 387.86 372.39 215.37 .555
Oct 1991 392.46 378.61 217.92 .555
Nov 1991 375.22 381.74 208.35 .555
Dec 1991 417.09 386.74 208.35 .500
Jan 1992 408.79 390.10 204.20 .500
Feb 1992 412.70 393.84 206.15 .500
Mar 1992 403.69 395.22 201.65 .500
Apr 1992 414.95 399.60 207.28 .500
May 1992 415.35 402.35 207.48 .500
Jun 1992 408.14 403.63 203.88 .500
Jul 1992 424.21 407.26 211.90 .500
Aug 1992 414.03 409.42 206.82 .500
Sep 1992 417.80 413.68 208.70 .500
Oct 1992 418.68 413.83 209.14 .500
Nov 1992 431.35 416.09 215.47 .500
Dec 1992 435.71 418.39 217.65 .500
Jan 1993 438.78 421.90 219.18 .500
Feb 1993 443.38 424.74 221.48 .500
Mar 1993 451.67 428.38 225.62 .500
Apr 1993 440.19 431.58 219.89 .500
May 1993 450.19 434.18 224.88 .500
Jun 1993 450.53 437.83 225.05 .500
Jul 1993 448.13 440.86 223.85 .500
Aug 1993 463.56 445.35 231.56 .500
Sep 1993 458.93 448.11 229.25 .500
Oct 1993 467.83 451.32 233.69 .500
Nov 1993 461.79 453.62 230.68 .500
Dec 1993 466.45 455.93 233.00 .500
Jan 1994 481.61 458.92 240.58 .500
Feb 1994 467.14 461.62 233.35 .500
Mar 1994 445.77 461.17 222.67 .500
Apr 1994 450.91 461.21 222.67 .494
May 1994 456.50 462.05 222.67 .488
Jun 1994 444.27 460.12 222.67 .501
Jul 1994 458.26 460.05 222.67 .486
Aug 1994 475.49 460.82 222.67 .468
Sep 1994 462.69 460.91 216.68 .468
Oct 1994 472.35 461.50 221.20 .468
Nov 1994 453.69 458.71 212.46 .468
Dec 1994 459.27 457.92 212.46 .463
Jan 1995 470.42 460.39 217.62 .463
Feb 1995 487.39 464.03 225.47 .463
Mar 1995 500.71 468.45 231.64 .463
Apr 1995 514.71 475.50 238.11 .463
May 1995 533.40 483.01 246.76 .463
Jun 1995 544.75 489.94 252.01 .463
Jul 1995 562.06 499.88 260.02 .463
Aug 1995 561.88 508.83 259.93 .463
Sep 1995 584.41 521.90 270.36 .463
Oct 1995 581.50 534.12 269.01 .463
Nov 1995 605.37 547.62 280.05 .463
Dec 1995 615.93 560.47 284.94 .463
Jan 1996 636.02 574.00 294.23 .463
Feb 1996 640.43 586.58 296.27 .463
Mar 1996 645.50 597.79 298.62 .463
Apr 1996 654.17 608.73 302.63 .463
May 1996 669.12 619.43 309.54 .463
Jun 1996 670.63 630.31 310.24 .463
Jul 1996 639.95 635.86 296.05 .463
Aug 1996 651.99 642.91 301.62 .463
Sep 1996 687.31 651.11 317.96 .463
Oct 1996 705.27 660.04 326.27 .463
Nov 1996 757.02 672.14 350.21 .463
Dec 1996 740.74 682.17 342.68 .463
Jan 1997 786.16 696.24 363.69 .463
Feb 1997 790.82 709.90 365.84 .463
Mar 1997 757.12 718.70 350.25 .463
Apr 1997 801.34 731.77 370.71 .463
May 1997 848.28 752.61 392.43 .463
Jun 1997 885.14 775.92 409.48 .463
Jul 1997 954.29 802.62 441.47 .463
Aug 1997 899.47 822.04 416.11 .463
Sep 1997 947.28 841.06 438.22 .463
Oct 1997 914.62 858.45 423.12 .463
Nov 1997 955.40 875.38 441.98 .463
Dec 1997 970.43 893.34 448.93 .463
Jan 1998 980.28 915.65 453.49 .463
Feb 1998 1,049.34 940.45 485.44 .463
Mar 1998 1,101.75 965.80 509.68 .463
Apr 1998 1,111.75 988.46 514.31 .463
May 1998 1,090.82 1,002.11 504.63 .463
Jun 1998 1,133.84 1,025.55 524.53 .463
Jul 1998 1,120.67 1,042.89 518.44 .463
Aug 1998 957.28 1,047.16 442.85 .463
Sep 1998 1,017.01 1,053.32 442.85 .435
Oct 1998 1,098.67 1,066.14 442.85 .403
Nov 1998 1,163.63 1,084.48 469.03 .403
Dec 1998 1,229.23 1,102.47 495.48 .403
Jan 1999 1,279.64 1,120.25 515.80 .403
Feb 1999 1,238.33 1,132.91 499.14 .403
Mar 1999 1,286.37 1,152.47 518.51 .403
Apr 1999 1,335.18 1,172.60 538.18 .403
May 1999 1,301.84 1,190.72 524.74 .403
Jun 1999 1,372.71 1,232.26 553.31 .403
Jul 1999 1,328.72 1,263.43 535.58 .403
Aug 1999 1,320.41 1,285.61 532.23 .403
Sep 1999 1,282.71 1,297.51 517.03 .403
Oct 1999 1,362.93 1,310.88 517.03 .379


Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241578 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 6:48 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
dackle7 (241577)

Over the last 5years to date (13 February 2013) my Sell to Buy system would have produced the following performance:


Feb 2008 107.09
Mar 2008 101.38
Apr 2008 112.96
May 2008 115.76
Jun 2008 110.61
Jul 2008 102.62
Aug 2008 106.03
Sep 2008 90.14
Oct 2008 73.01
Nov 2008 90.66
Dec 2008 96.73
Jan 2009 97.68
Feb 2009 101.55
Mar 2009 138.38
Apr 2009 159.44
May 2009 179.38
Jun 2009 176.09
Jul 2009 195.98
Aug 2009 213.64
Sep 2009 242.18
Oct 2009 227.39
Nov 2009 244.60
Dec 2009 255.93
Jan 2010 247.77
Feb 2010 260.35
Mar 2010 283.75
Apr 2010 286.74
May 2010 273.32
Jun 2010 257.16
Jul 2010 289.73
Aug 2010 299.08
Sep 2010 305.65
Sep 2010 337.60
Oct 2010 326.35
Nov 2010 331.41
Dec 2010 363.30
Jan 2011 375.33
Feb 2011 406.43
Mar 2011 383.77
Apr 2011 410.00
May 2011 390.85
Jun 2011 398.76
Jul 2011 399.84
Aug 2011 365.33
Sep 2011 319.96
Oct 2011 358.85
Nov 2011 341.36
Dec 2011 342.09
Jan 2012 372.53
Feb 2012 402.60
Mar 2012 412.80
Apr 2012 403.08
May 2012 349.33
Jun 2012 359.38
Jul 2012 364.10
Aug 2012 356.54
Sep 2012 394.54
Oct 2012 404.38
Nov 2012 410.87
Dec 2012 430.92
Jan 2013 452.15
Feb 2013 484.47


Can you provide the S&P data for the same period for comparison?

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: dackle7 One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241579 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 7:02 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
AliciaThyme, my post was more of a friendly little jab at RayVT, who appears to be a firm believer in Faber's 10-month moving average method.

BTW do you have any other details of your strategy, such as how you select candidate stocks?

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241580 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 10:00 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
Consider the results for an investor following the 10-month SMA strategy from 1982 to 1999. Buy-and-hold would've turned $10,709 into $136,293 plus dividends, whereas the Faber follower would've ended with only $51,703 plus dividends, but less taxes and commissions.

No, this is way off. (BTW, a minor quibble: both approaches would have had taxes - one along the way and the other all at the end.)

I get the following statistics.
SPX, without dividends:
SPX from 07/01/82 thru 12/01/99
B&H: $1.0000 grows to $12.9839 in 17.42 years. rate: 15.86% stdev: 15.8%
SMA: $1.0000 grows to $ 8.5473 in 17.42 years. rate: 13.11% stdev: 12.6%
SMA does 19 trades in 17 years, so commission is negligible.

But I see what you did there. You picked a time period where B&H did extremely well. We essentially agree on the B&H return, but you get a lot lower return for SMA than I do.

Extending the end date to include the "lost decade":
SPX from 07/01/82 thru 12/20/12
B&H: $1.0000 grows to $13.4110 in 30.47 years. rate: 8.89% stdev: 18.4%
SMA: $1.0000 grows to $ 9.1835 in 30.47 years. rate: 7.55% stdev: 11.9%

And for just the "lost decade":
SPX from 07/01/99 thru 12/20/12
B&H: $1.0000 grows to $ 1.0377 in 13.47 years. rate: 0.28% stdev: 21.3%
SMA: $1.0000 grows to $ 1.0220 in 13.47 years. rate: 0.16% stdev: 10.8%

Both got essentially the same return. None. But SMA had only half the volatility, and half the MaxDrawDown.

and
SPX from 01/03/50 thru 12/20/12
B&H: $1.0000 grows to $85.0230 in 62.96 years. rate: 7.31% stdev: 15.5%
SMA: $1.0000 grows to $69.7556 in 62.96 years. rate: 6.97% stdev: 10.2%
SMA does 95 trades in 63 years.

Faber's SMA has never been about increasing the absolute return. It's been about getting almost the same return with much lower volatility. IOW, about inproved risk-adjusted return.

Of course, you cannot buy SPX. You can only buy an index fund -- which has tracking error and taxes due to internal trading and expense ratio.

With dividends (can only go back to 1987):
VFINX from 03/27/87 thru 12/01/99
B&H: $1.0000 grows to $ 6.4259 in 12.68 years. rate: 15.80% stdev: 16.7%
SMA: $1.0000 grows to $ 4.4582 in 12.68 years. rate: 12.51% stdev: 12.5%

and
VFINX from 03/27/87 thru 10/07/11
B&H: $1.0000 grows to $ 6.4941 in 24.53 years. rate: 7.92% stdev: 19.4%
SMA: $1.0000 grows to $ 5.3239 in 24.53 years. rate: 7.05% stdev: 11.7%

Now here's where it gets interesting. The lost decade *with* dividends:
VFINX from 07/01/99 thru 10/07/11
B&H: $1.0000 grows to $ 1.0207 in 12.27 years. rate: 0.17% stdev: 21.7%
SMA: $1.0000 grows to $ 1.1920 in 12.27 years. rate: 1.44% stdev: 10.7%

If we change the SELL signal to be 3% below the SMA:
VFINX from 03/27/87 thru 10/07/11
B&H: $1.0000 grows to $ 6.4941 in 24.53 years. rate: 7.92% stdev: 19.4%
SMA: $1.0000 grows to $ 6.5334 in 24.53 years. rate: 7.95% stdev: 12.5%

Now the SMA method gets the same return, but only 2/3 the volatility.


-----
As you might gather from the format of the above data, I have a program which reads historical price & dividend data files, applies trading rule parameters, and spits out the statistics. I'm pretty sure it is correct, but am always cognizant that it may have bugs.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241581 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 10:32 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
Over the last 5years to date (13 February 2013) my Sell to Buy system would have produced the following performance:
...
Feb 2013 484.47


Easy. SPY went from 123 to 152, which would be $100 grows to $123, vs. your ending value of $484.47. Or just pull up the monthly historical quotes for SPY from yahoo, to see the month-by-month values.

But what you have described isn't, strictly speaking, a system -- since you clearly have trades that don't correspond to the rules you stated.
Plus you haven't stated how to pick the stocks. And that's a key component of a system, picking the stocks.

BTW, the CAGR of your portfolio is 48%. A portfolio that is getting 48% CAGR when the broad market is getting 5.4% CAGR is taking a huge amount of risk.

"Stock market history is littered with people who misunderstood the risks they were taking on." -- William Bernstein

"[About Amaranth Capital:] The evidence of potential risk came primarily in the form of the outsized gains, and these are rarely recognized as the red flag they are."

Richard Feynman: "The first principle is that you must not fool yourself—and you are the easiest person to fool."


You are going to ignore me about this.
:shrug: That's okay.
Maybe you should print this out and tuck it away to re-read in 5 years. Or whenever your portfolio gets crushed before then. And, who knows---maybe in the next 5 years you'll continue to get 48% annual return and your portolio is worth a bazillion dollars. Then you can re-post this message here and laugh in my face.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: dackle7 One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241582 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 11:11 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
I rechecked my calcs and everything appears to be in order. The dates are the last trading day of each month, so July 30, 1982 (S&P = 107.09) to October 28, 1999 (S&P = 1362.93). If the S&P is above its 10-month moving average on the last day of the month, the position is held until the last day of the following month.

The 10-month MA strategy did very well from 1969-82, and then got clobbered from 1982-99. Then it did well again from 1999-2009, and then once again it is lagging (eg GTAA ETF is up 0.3% since inception vs 28.6% for SPY -- ouch!).

I think it is important to keep in mind that the two bear markets over the last 13 years are somewhat anomalous, and that they make moving average strategies look a bit better than they really are.

Print the post Back To Top
Author: musselmant Three stars, 500 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241583 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 11:12 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Screen S&P

CAGR 10 10
GSD 7 18
Sharpe 0.64 0.32

6912BLtb11p75rs120p25 since 1969

75% Tbills
25% RS26 1-120

And no market timing needed.

Print the post Back To Top
Author: musselmant Three stars, 500 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241584 of 253091
Subject: Re: Does this stuff work ? Date: 2/13/2013 11:23 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
 formatting and typos fixed:

Screen S&P

CAGR 10 10
GSD 7 18
Sharpe 0.64 0.32

http://backtest.org/6912BLtb11p75rs120p25 From 1969 forward thru 2012

75% Tbills
25% RS26 1-20

And no market timing needed.




Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241585 of 253091
Subject: Re: Does this stuff work ? Date: 2/14/2013 9:26 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Rayvt (241581)

Maybe you should print this out and tuck it away to re-read in 5 years. Or whenever your portfolio gets crushed before then. And, who knows---maybe in the next 5 years you'll continue to get 48% annual return and your portolio is worth a bazillion dollars. Then you can re-post this message here and laugh in my face.

I have been there already.

My system was developed from historical data 5 years ago as I explained earlier.
For the record, the results for a 5year portfolio ending March 2008 was as follows:


Mar 2003 107.09
Apr 2003 113.83
May 2003 122.36
Jun 2003 140.84
Jul 2003 137.88
Aug 2003 151.62
Sep 2003 163.22
Oct 2003 159.64
Nov 2003 160.67
Dec 2003 161.39
Jan 2004 171.80
Feb 2004 178.27
Mar 2004 165.41
Apr 2004 170.80
May 2004 158.37
Jun 2004 156.81
Jul 2004 154.03
Aug 2004 150.28
Sep 2004 164.78
Oct 2004 172.66
Nov 2004 182.90
Dec 2004 188.67
Jan 2005 203.04
Feb 2005 211.79
Mar 2005 218.05
Apr 2005 213.30
May 2005 202.75
Jun 2005 204.30
Jul 2005 231.84
Aug 2005 253.47
Sep 2005 271.06
Oct 2005 267.67
Nov 2005 282.14
Dec 2005 303.60
Jan 2006 336.45
Feb 2006 357.68
Mar 2006 353.39
Apr 2006 382.36
May 2006 386.08
Jun 2006 367.13
Jul 2006 377.76
Aug 2006 386.07
Sep 2006 403.43
Oct 2006 406.61
Nov 2006 454.57
Dec 2006 452.58
Jan 2007 454.35
Feb 2007 451.70
Mar 2007 451.91
Apr 2007 479.74
May 2007 504.24
Jun 2007 501.46
Jul 2007 530.26
Aug 2007 499.97
Sep 2007 531.82
Oct 2007 562.54
Nov 2007 536.04
Dec 2007 566.89
Jan 2008 529.32
Feb 2008 545.69
Mar 2008 509.94


That was the reason for committing to the system.

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241586 of 253091
Subject: Re: Does this stuff work ? Date: 2/14/2013 9:32 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
dackle7 (241579)


BTW do you have any other details of your strategy, such as how you select candidate stocks?

Having been hammered by disclosure on the UK board, my selection system is no longer up for discussion.
However, I am working on a variant which will select one stock every 10 days (or maybe every month) based on selecting the stock having the best recovery during the previous month. The recovery will be assessed by measuring the acceleration of the rising trend (Excel TREND function).

If it works I will post details.

Alicia

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241588 of 253091
Subject: Re: Does this stuff work ? Date: 2/14/2013 6:24 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
BTW do you have any other details of your strategy, such as how you select candidate stocks?

Having been hammered by disclosure on the UK board, my selection system is no longer up for discussion.
However, I am working on a variant which will select one stock every 10 days (or maybe every month) based on selecting the stock having the best...


I managed to find that thread. http://boards.fool.co.uk/low-risk-investment-strategy-127306...

Wow! So far, 204 messages in the thread. That's gotta be some sort of record.
Sitting down to read it now. Trying to decide if I should go fix some popcorn first. ;-)

Print the post Back To Top
Author: FlyingCircus Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241608 of 253091
Subject: Re: Does this stuff work ? Date: 2/16/2013 1:06 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Well, I did just scan most of that thread, and, not surprisingly, it evolved much the way this one is.

This board's primary method of selling rule has been easy: after a set time - one month, 3 weeks, one quarter, six months, one year.
Hold till drop is a modification.
Profit latch is another modification.
Stop loss is yet a fourth modification.

But the vast majority of focus here, has always been on the buy selection criteria.

AT's proposal is switching these priorities. essentially, it doesn't matter what the buy selection criteria are (assuming there are many out there that can identify attractive candidates, i.e. the 250+ MI screens). It's the sell criteria that becomes more complicated - as posted extensively.

To backtest this in this world, one would have to
1.) generate the stock picks for one given screen or a blend, whatever your choice is, going back x# of years on a y frequency
2.) download the daily prices of each of those picks
3.) build a model tracking the returns by position by week, and then make the sell/switch determination by the current results of the position.

Our backtesters manage steps 1 and 2. The third step would only be possible at GTR1 and I'm pretty sure a complex exercise even if Robbie built the calculations.

Lot of work to test this.

FC

Print the post Back To Top
Author: FlyingCircus Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241612 of 253091
Subject: Re: Does this stuff work ? Date: 2/16/2013 3:58 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
So - after brute forcing my way for 2 hours through a manual backtest running this on RS26 for the last year, somehow I realized it's simply a 5% profit latch system, with a Hold-till-it-comes back-unless it melts-down-40%

So this was on one screen. Starting a year ago (early February) through November before I stopped consuming time on this, it ended up being up about 20% - after being down 20% in a few short months.

The weekly-traded RS26 was up 9.8% the last year by comparison.

Anecdotal only, short backtest. I maintain the success of the system is the ability of the buy criteria/screen to identify stocks that have a high probability of going up 5% in the next n weeks/months than down.

Not sure why I couldn't figure that out first instead of burning 2 hours in a show me exercise.... maybe because this cold is frying my brain...

FC

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241670 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 3:32 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
FlyingCircus (241612)

Glad to see someone is taking the system seriously.

I maintain the success of the system is the ability of the buy criteria/screen to identify stocks that have a high probability of going up 5% in the next n weeks/months than down.

I may be naive, but isn't that what all buy selection criteria are supposed to do - I recognise that even the professionals don't always get it right, but surely they are best placed to make good judgements.

I am now working on the new buying selection method which will allow anyone using a spreadsheet to backtest my sell-to-buy scheme for themselves. It's looking promising, up there with the best on 5 year-to-date share prices (about 118% net profit), but needs more refining because some 5year periods within the 10-year dataset are not so good.

I will also run it on the Dow Jones dataset for the last 10 years.
Does anyone know of a free downloadable source for dividends of DJ constituents?

Alicia.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241671 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 3:45 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
To help with my model for DJ constituents I need to know what brokerage charges you guys pay in US. (in UK we pay approx $8 to buy and $8 to sell, regardless of quantity, plus when buying an additional 0.5% in tax to the government) I take these costs into account to get the net profit result.

Alicia

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241672 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 3:53 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Does anyone know of a free downloadable source for dividends of DJ constituents?

Correction - I know I can find dividends at Yahoo, but I need ex-dividend dates as well as pay dates.

Thanks

Alicia

Print the post Back To Top
Author: warrl Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241675 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 4:36 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
I maintain the success of the system is the ability of the buy criteria/screen to identify stocks that have a high probability of going up 5% in the next n weeks/months than down.

I may be naive, but isn't that what all buy selection criteria are supposed to do


Not necessarily. I recently picked up a stock on a tip from a person who was scared off by the exact feature I like: the frequent one-day 10% swings in *both* directions. I don't have much faith in the stock long-term and will be checking frequently to see if it's time to unload the position completely, but in the meantime I'm doing a mechanical volatility play on it.

If a stock is not likely to go up 5% *and* go down 5%, I am not really interested.

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241676 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 4:55 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
I know I can find dividends at Yahoo, but I need ex-dividend dates as well as pay dates.
The pay dates don't matter. What matters is the ex-div date --- and what yahoo gives in the historical data downloads is the ex-div date.

Commissions are all over the map, from free to huge. $8 to $10 is a good estimate.

I am now working on the new buying selection method which will allow anyone using a spreadsheet to backtest my sell-to-buy scheme for themselves.

None of this matters, though. Nobody could try backtesting your sell-to-buy system in their own portfolios, because nobody puts 1/3 of their money in each of 3 stocks, and then waits for one of them to gain 5% and then sells it to buy another stock.

Print the post Back To Top
Author: warrl Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241677 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 5:00 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
To help with my model for DJ constituents I need to know what brokerage charges you guys pay in US. (in UK we pay approx $8 to buy and $8 to sell, regardless of quantity, plus when buying an additional 0.5% in tax to the government)

At InteractiveBrokers.com for US customers and stuff traded in US and Canadian markets, as a rough rule of thumb figure $1 for up to 200 shares plus half a cent per additional share for stocks, and $1.07 per contract (100 shares) for options. That's for both purchases and sales (including short-selling, at least on covered call options).

There are various circumstances in which they charge lower commissions; I've been there fairly often on options (usually writing covered call options on SLW) but never yet on stocks.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241678 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 5:11 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Ravvt (241676)

the pay dates don't matter. What matters is the ex-div date --- and what yahoo gives in the historical data downloads is the ex-div date.

The pay dates matter to me because I add them into the accumulated funds for reallocation to a new target buy.

nobody puts 1/3 of their money in each of 3 stocks, and then waits for one of them to gain 5% and then sells it to buy another stock.

I guess you are right about that - but isn't anyone curious about how to improve their investment returns?

If not then I won't bother you further.
Please let me know.

Alicia

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241679 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 5:23 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 24
None of this matters, though. Nobody could try backtesting your sell-to-buy system in their own portfolios, because nobody puts 1/3 of their money in each of 3 stocks, and then waits for one of them to gain 5% and then sells it to buy another stock.


Sheesh Ray, who are you speaking for? If you don't like that exact method you could, for example, start out with 10 stocks and then follow the method. Or you could ignore it completely. But quit throwing cold blankets on ideas just because they don't exactly match YOUR investing style.

Elan

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241680 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 5:44 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
Yup, shoulda made some popcord.
Very interesting thread(s). The Brits are *very* funny -- I almost peed my pants a couple of times.

This is the 3rd thread on 3 different message boards/forums (that I've found - there may be more) where Alicia has presented this idea. Always with negative feedback. Always with the same overall objections & critique.

A brit poster said, "I have reached the conclusion that you [Alicia] are either having us on or are deluded." Personally, I've come to the conclusion that Alicia is deluded about the strategy and obstinately refuses to see any risk or shortcomings.

In addition to dozens of people asking pointed questions, there have also been several posters offering cogent analysis and critique. All of which have been casually dismissed or ignored by you.

For example, the rules say to sell a stock only if it is showing a 5% profit or a 40% loss. Your examples/historical trades always show that one of these triggers is hit. But a number of people have pointed out that stocks can stay inside these limits and *never* hit a trigger, so you wind up holding dogs that either never go up or go down a little and stay there. If that happens, you can never buy anythinhg else. But the instant that a stock hits a 5% gain you immediately sell it to buy a new pick. So over time the portfolio will be full of dogs -- because you keep the ones that go nowhere and sell the good ones as soon as they have a little profit.

Still waiting for you to respond to this criticism -- or even to acknowledge that this scenario could happen.

The rules also tell you to buy a stock that has gone down 50% if you EVER had gotten a buy signal for it. Even if the buy signal was 3 years ago! One stock such as Eastman Kodak would have wiped out your portfolio. The all-time high was 80. So you would have bought it at 40. Then sold it at 24 (when it hit 40% loss). Then bought it at 20 when it hit 50% loss. And so on and so on -- all the way down to its current price of 0.20. Each time you bought, you would have sold your best holding as the source of funds. Your entire portfolio would have gone down this one rathole.

I now hope someone will take me seriously, try the scheme on their own holdings, and let me know their conclusion.

Several people have told you their conclusions -- that selling their best performing stock to buy a new stock is an inferior strategy. And every time, you ignore what they say. The only rationale you give for your method is "It worked for me."

Frankly, I don't even believe it worked for you. I wouldn't believe the results you claim unless I saw your complete official trading records --- which I am not asking to see, mind you, just saying that that's the only thing that would make me believe your claim.

It is interesting that you believe that the most important part of your method is the "sell-to-buy" part (that is, sell the stock that has the most profit and only sell a stock that has 5% or more profit) and that the least important part is the stock picking part.

Everybody else -- EVERYBODY ELSE -- thinks that the most important part is the stock picking part, and that the "sell-to-buy" part is either unimportant or is actually an overall detriment.

That's why somebody always says that your strategy sounds like a Martingale system. Because the key theory of Martingale is that you can shift the odds in your favor by changing the mechanics of how you bet.


You always say that this is a low-risk strategy. Yet everybody who has commented about the risk aspect has said, "No, this is a high-risk strategy." Nobody else has agreed that it is low risk. Dozens of experienced investors who commented to you have said this is not low risk, yet you -- an apparent neophyte -- still insist it is low risk.
Does that not give you pause? Does that not ring any alarm bells that perhaps you should think more carefully about it?

This whole strategy is a conglomeration of bad ideas. Things that are widely known to be detrimental to investment success.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241681 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 5:53 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
- but isn't anyone curious about how to improve their investment returns?

Of course we are. Everybody here is interested to improve their investment returns. And a big part of that is to avoid doing things that *hurt* your returns.

We can do simple math, and we can figure out that it takes eight (8) 5% profits to pay for one 40% loss.

The reason that these message boards are valuable is that we share ideas that are good and also ideas that are bad. Investment success comes from doing the good ideas and NOT doing the bad ideas.

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241682 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 6:11 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
If you don't like that exact method you could, for example, start out with 10 stocks and then follow the method. Or you could ignore it completely. But quit throwing cold blankets on ideas just because they don't exactly match YOUR investing style.

Work it through, Elan, work it through.

Actually, the bit about using more that 3 stocks got discussed quite a bit on the 2 Brit forums. Their conclusion was that while the risk is reduced by having more stocks, the strategy would work worse.

The tactic of selling a stock as soon as it is showing a small (5%) profit has been known to be a bad tactic ever since, well, forever.

Richardo wrote about that in 1805.

”—2) Cut short your losses,— 3) Let your profits run on.” By cutting short one’s losses, Mr. Ricardo meant that when a member had made a purchase of stock, and prices were falling, he ought to resell immediately. And by letting one’s profits run on he meant, that when a member possessed stock, and prices were rising, he ought not to sell until prices had reached their highest, and were beginning again to fall."

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241683 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 6:21 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Rayvt (241680)

NO ONE has yet PROVED that my strategy is flawed.
I have proved that it works.

Would you have me give up because others disagree with me?

Yes, there has been a tirade of criticism - but these are by people who seem unable to consider that there MAY be other ways than conventional trading
methods.

There have also been a few favourable comments.

I am doing my best to convert my system into something you could model for yourself to get your own proof.

I am looking at trend acceleration, by which I mean the difference between a trend over the last 20 days and the trend over the previous 20 days. By calculating this value each day for each stock the target buy will be the stock with the highest result. For example with XOM in a 40 row Excel spread sheet showing columns A:E >>


30/11/2012 88.14
03/12/2012 87.61
04/12/2012 87.19
05/12/2012 87.73
06/12/2012 88
07/12/2012 88.6
10/12/2012 88.41
11/12/2012 88.99
12/12/2012 89.45
13/12/2012 88.58
14/12/2012 88.08
17/12/2012 88.87
18/12/2012 89.56
19/12/2012 88.44
20/12/2012 88.89
21/12/2012 87.23
24/12/2012 86.92
26/12/2012 87.07
27/12/2012 86.86
28/12/2012 85.1
31/12/2012 86.55
02/01/2013 88.71
03/01/2013 88.55
04/01/2013 88.96
07/01/2013 87.93
08/01/2013 88.48
09/01/2013 88.14
10/01/2013 89.1
11/01/2013 89.61
14/01/2013 89.58
15/01/2013 89.53
16/01/2013 89.47
17/01/2013 90.2
18/01/2013 90.8
22/01/2013 90.92
23/01/2013 90.7
24/01/2013 91.35
25/01/2013 91.73
28/01/2013 91.11 C40 D40 E40
29/01/2013 91.76 -0.08% 0.24% 0.32%

C40 = TREND(B1:B20,,{21})/TREND(B1:B20,,{20})-1
D40 = TREND(B21:B40,,{21})/TREND(B21:B40,,{20})-1
E 40 = D40-C40



The result of 0.32% (cell E40) is logged for each of the 30 DJ stocks.
The stock having the highest result becomes the target buy for that day.
The next target buy could be 10 days later, and you then apply my Sell-to-Buy rules.


This is where I am at the moment, subject to development, and anyone can run it if they wish.

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: OneStepBeyond Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241684 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 6:21 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 9
We can do simple math, and we can figure out that it takes eight (8) 5% profits to pay for one 40% loss.

Are you deliberately mischaracterizing her system? She has said she will only sell at a 5% profit if there is another stock she wants to buy, if not, the other three stocks can appreciate beyond 5%.

It sounds like she is trying to illustrate her system with a spreadsheet and make it available to anyone interested. IMO, that is a more productive attitude than slagging her system before she gets a chance to demonstrate it.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241685 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 6:25 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Rayvt (241682)

The tactic of selling a stock as soon as it is showing a small (5%) profit has been known to be a bad tactic ever since, well, forever.

I agree.
Read my rules

Alicia

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241686 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 6:28 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Work it through, Elan, work it through.

No thanks, I'm not going to work it through because it doesn't seem to me like a good system. However, since I'm not going to work it, I'm also not going to throw darts at it.

I think the biggest potential flaw is the sell at 5% gain decision, even when the stock could very well continue to qualify as a buying candidate.

Elan

Print the post Back To Top
Author: rharmelink Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241687 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 7:05 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
I know I can find dividends at Yahoo, but I need ex-dividend dates as well as pay dates.

Try earnings.com:

http://www.earnings.com/company.asp?client=cb&ticker=mmm...

Print the post Back To Top
Author: DrBob2 Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241688 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 7:46 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Everybody else -- EVERYBODY ELSE -- thinks that the most important part is the stock picking part, and that the "sell-to-buy" part is either unimportant or is actually an overall detriment.

Well, not everybody. In one of the Wizard interviews (and I don't remember which one) the trader said that with proper position management and rules he could do very well with a random selection of stocks. What you did with the hand was more important than the hand dealt.

Of course, you have to have a good set of rules....

DB2

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241692 of 253091
Subject: Re: Does this stuff work ? Date: 2/18/2013 11:37 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 3
Ah well, I get to play Devil's Advocate again.
"The purpose of such process [Devil's Advocate] is typically to test the quality of the original argument and identify weaknesses in its structure, and to use such information to either improve or abandon the original position."

Are you deliberately mischaracterizing her system? She has said she will only sell at a 5% profit if there is another stock she wants to buy, if not, the other three stocks can appreciate beyond 5%.

On the Brit TMF board, she said that she had an automatically running program that downloaded quotes and ran signals in the middle of every afternoon, and immediately emailed and texted the signals to her so she could trade immediately, before the market closed. She also said that most of the time the signal was "No ACtion". She said the universe was the FTSE 100 stocks, so of course most of the signals would be no action. She also said (with respect to buying a stock that hit a 50% loss) that she remembered the candidate picks that were bypassed due to lack-of-funds for as long as 3+ years. It was also stated that many buy signals were bypassed because there was no candidate "source of funds" stock to buy, but that the signal would be remembered and acted upon when a held stock hit a sell trigger.

If a buy signal was very infrequent, then there is no need to be in a rush to buy on an intra-day signal in mid-afternoon.

It was also stated that you start out buying 3 stocks, each with 1/3 of your money, and you would buy the 4th stock as soon as you had funds to do so. She talked about selling a stock, using the proceeds to buy the original dollar amount of a new stock, and bank the excess. When these accumulated banked excesses become equal to the original dollar amount, you buy another new stock thereby increasing your total number of stocks by 1.

It was also stated: "The purpose is to take profit when available at the time you want to buy another, so stock#4 is bought from the proceeds of selling the best of the three holdings (all stock purchases are made to the same purchase value)."

The implication of all the above is that there is almost always an active buy candidate waiting in the wings, and that as soon as a held stock hits 5% profit (or 40% loss), then it is sold and the funds used to buy a candidate stock. (Left unsaid is which stock to buy, if there are several un-bought candidates.)

Posters on the Brit board had already worked that out, and concluded that what would happen in practice was that as soon as a stock hit 5% profit it would almost always be sold. IOW, there is almost always a candidate stock to buy, so it's just waiting for an existing held stock to hit its sell trigger.

Now, the wierd part is, [again re-stating what posters on the Brit board said] that if you are holding 3 stocks with unrecognized gains of: 6%, 5%, and -20%, you sell the one with the 6% stock to buy the new stock. Now your unrecognized gains are: 5%, -20%, and 0%. If you have another buy candidate, you sell the 5% stock, and now your portfolio stocks area at -20%, 0%, and 0%.

You've sold your best winners and kept all the losers. That -20% loser will *stay* in your portfolio. One-third of your portfolio is cemented in to a loser stock.
Once you get another dog that loses, say -10%, then you have TWO stocks that you keep and never sell. The portfolio is now 67% dreck.

One of the Brits pointed out that if your initial 3 buys stay flat, the system stalls and you never, ever sell any of them or buy a new stock. While this is perhaps unlikely, it *is* a distinct possibility and therefore must be addressed. This failure mode must be addressed. A strategy of "only the circumstances that I want to consider and all others will be ignored" doesn't cut the mustard,

It sounds like she is trying to illustrate her system with a spreadsheet and make it available to anyone interested. IMO, that is a more productive attitude than slagging her system before she gets a chance to demonstrate it.

I am looking forward to this spreadsheet.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241693 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 12:26 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
NO ONE has yet PROVED that my strategy is flawed.
The scientific method doesn't work that way. It works the opposite way.

There are a couple of theory failure modes that we have to consider. There are Type 1 and Type 2 errors.

For investing theories, these would be:
1) Think that the strategy doesn't work when it actually does.
2) Think that the strategy works when it actually does not.

Then you apply the question, "But what if I am wrong?"

For the 1st type of error, you don't use a strategy that would make you a profit, so you miss out on a gain.

For the 2nd type of error, you use a strategy that is a failure, so you lose money.

This 2nd type of error is fatal, since you lose your money.

The other possibilities aren't errors, and are:
3) Think that the strategy doesn't work and it actually does not.
4) Think that the strategy works and it actually does.

#3 is good, because you accurately predict that it will lose your money, and you don't do that.

#4 is the best of all, because you accurately predict that it will make you a PROFIT.


I have proved that it works.
Well, no you haven't. At best, you have convinced yourself that in looking back you would have made a profit.

The problem is that your strategy violates a number of precepts that are well known and commonly accepted as BAD things that lose money instead of making a profit. The first thing that needs to be done is to explain why action X leads to good outcomes in your system even though X is generally known to be bad.

For example, your system says to sell winners and keep losers.
Another example is buying (or re-buying) a stock if it hits a 50% loss.
The generally accepted wisdom is to keep winners and sell losers, and to avoid a stock that has lost 50%.

trend acceleration, by which I mean the difference between a trend over the last 20 days and the trend over the previous 20 days.
Okay. This is a variant of momentum, which is a well known technique. But you earlier said that it was this *and* the stock had dropped in the last day. Which is it?

You will get the question -- and you should have already asked this question yourself in developing the system-- how sensitive is that number 20? What happens if you use 19? or 18? Or 21? Or 22?

TREND is the Excel function for a least-squares fit. A more common method for determining momentum trends is just looking at the end-points, i.e., the prices 20 days apart and ignoring the intermediate prices. Why did you use TREND instead of the standard common method? What is the justification for it? (And "because it works" is not a justification.)
What happens if you use the common method instead of TREND?

You have to ask sharp questions of every part of a strategy, to make sure that none of the steps make an error Type 2 -- because those are the errors that will lose all your money.

Are you wanting to feel good, or are you wanting to make money? Me, I want to make money -- If I propose some idea and somebody says "That's a stupid idea that will lose money" -- that's just the type of feedback I want.

-------
And this entire discussion belongs in its own thread.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241696 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 1:16 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 26
I have proved that it works.
Well, no you haven't. At best, you have convinced yourself that in looking back you would have made a profit.


She has written that the system has made real money with real investing for the last five years. It could be another case of the Beardstown Ladies, but you gotta give a person the benefit of the doubt on such a claim, unless you're prepared to audit their returns.

Having said that, all your reasoning about why it shouldn't work isn't worth squat. You and I and everyone here is about the proof of the pudding, not about critiquing the recipe for the pudding. In plain English, the backtest is the only thing that counts. She claims to have a 5 year backtest. As far as I can tell it's not a repeatable backtest, it's not long enough, and it may not be statistically robust. Rather than waving our arms about why it shouldn't work, the constructive advice should be about how to turn it into a testable, repeatable strategy. Then it can be put to a valid backtest.

Elan

Print the post Back To Top
Author: lcrisorio One star, 50 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241698 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 8:22 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
RE:--"does anyone want to improve their investment returns"? Anyone who doesn't reply with a very strong "YES" is not being "real".
Please proceed.
lcrisorio

Print the post Back To Top
Author: gritton Big red star, 1000 posts Top Favorite Fools Feste Award Winner! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241707 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 10:47 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
RE:--"does anyone want to improve their investment returns"? Anyone who doesn't reply with a very strong "YES" is not being "real".

Anyone who doesn't reply with a roll of the eyes and a sigh is ... well, someone I can't relate to.

- Jamie

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241711 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 11:14 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Rather than waving our arms about why it shouldn't work, the constructive advice should be about how to turn it into a testable, repeatable strategy. Then it can be put to a valid backtest.

Yeah, this morning I actually was thinking about how I might go about doing that.

First you'd have to pick a universe of stocks. IIRC, she used the FTSE100, which is no good for most of us American investors. The next candidates would probably be the 30 DJIA stocks or the S&P500 stocks. DJ stocks are questionable, but arguable okay. The 500 stocks of the S&P is a huge amount of data, probably unmanageably large. You need to keep track of the daily prices of each stock for 3+ years (for a real-life portfolio.

For a backtest, you need prices going back 10+ years.
That's 2500 prices for each stock, or 1,250,000 prices for all 500 stocks of the S&P500. You also need to keep track of dividends, at 4 per year per stock that's 20,000 dividends. The data alone is probably around a 10MB file.

Then you'd have Excel computing the formulas -- 4 TREND functions for each day (except the first 20 days). TREND is a least-squares fit, which is computationally complex. So doing 10,000 of these computations for each stock is going to take a long time.

You have to make sure you don't have survivorship bias in your dataset. You can't just use the stocks that are in the index today, you have to use the stocks that were in the index at the time for each step of the backtest. That adds considerable complication.

I guess you could simplify this by arbitrarily taking a random selection of 100 stocks from the S&P500 as of Jan 2000 and declaring that as your universe. You'd still have the problem of getting historical prices on any of those stocks that subsequenctly went bankrupt or bought out or merged or split up.

=== WHEW! === That's just the data.

Now the rules.
You'd have to come up with the set of rules, written down concretely. Lots of cases have to be covered. If you get a buy signal, but there's nothing to sell, what do you do? Remember the signal? If so, for how long? If you get several simultaneous buy signals, what do you do? If you can't buy them all, which do you buy?

All in all, to backtest this seems like a huge job. Even so, I wouldn't be unwilling to do it if there was a decent possibility that it would be a successful strategy.

The problem there is that one of her rules says to sell the winners and keep the losers. And that violates a heuristic that has been known for hundreds of years.

As they say on Shark Tank (Dragons' Den for the Brits), "It's a huge effort and I can't how this would work. And for that reason, I'm out."

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241713 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 1:20 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 10
Hi Guys

I have run my system on the 30 Dow Jones constituents using closing prices downloaded from Yahoo for 5 years to 13 Feb 3013 (ie last Wednesday). During that period the DJI rose just 11.4% from 12552 on 13 Feb 2008 to 13982 on 13 Feb 2013. In the following table I have listed all the transactions which were actioned, and shown both Profit and Net Profit (the latter assuming buy and sell costs of $10 per transaction). I have used the trend formula I previously posted and selected a new target buy every 10 days. I started with the same investment value I have been using previously - ie $10,500 and applied my Sell-to-Buy rules (except that I have not used the sell-on-crash condition at this time). At the end the final value without costs was $16,277, representing a gross profit of $5,777 (55%) and a net profit of 46.4% This result would place the scheme in the top 10% of 2700 uk professional funds.

Interestingly, the early buys included a series of stoploss buys, but even so, the overall net profit is still 4x the DJI increase.

I have not yet added in the effect of dividends, but if the UK FTSE100 results are anything to go by, I would expect to almost double the final result.

The sceptics can easily check the dates and prices.


EPIC Gross Bought @ Sold @ held Net Profit
AA 17.9% 15 Feb 2008 35.72 13 May 2008 42.12 89 16.7%
MRK -40.5% 03 Mar 2008 44.06 09 Oct 2008 26.21 221 -41.2%
MSFT -40.1% 17 Mar 2008 28.3 25 Feb 2009 16.96 346 -40.8%
DIS -41.3% 13 May 2008 34.33 12 Nov 2008 20.16 184 -42.0%
AA -40.1% 14 Nov 2008 10.84 18 Feb 2009 6.49 97 -40.8%
BAC 45.8% 27 Feb 2009 3.95 13 Mar 2009 5.76 15 45.1%
BAC 91.3% 13 Mar 2009 5.76 13 Apr 2009 11.02 32 90.6%
BAC 21.5% 27 Mar 2009 7.34 27 Apr 2009 8.92 32 20.8%
UNH 11.6% 13 Apr 2009 24.37 26 May 2009 27.19 44 10.9%
AXP 11.1% 27 Apr 2009 24.23 09 Jun 2009 26.93 44 10.4%
DIS 10.7% 11 May 2009 24.71 01 Oct 2009 27.36 144 10.0%
MRK 5.6% 26 May 2009 26.56 08 Jul 2009 28.04 44 4.9%
TEV 8.9% 09 Jun 2009 44.04 19 Aug 2009 47.96 72 8.2%
PFE 8.6% 08 Jul 2009 14.61 22 Jul 2009 15.87 15 7.9%
DD 12.7% 22 Jul 2009 28.47 05 Aug 2009 32.08 15 12.0%
CAT 15.5% 05 Aug 2009 46.64 17 Sep 2009 53.89 44 14.8%
BAC 8.1% 19 Aug 2009 16.75 15 Oct 2009 18.1 58 7.3%
PG 10.4% 17 Sep 2009 55.53 12 Nov 2009 61.3 57 9.7%
GE 16.0% 01 Oct 2009 15.97 09 Apr 2010 18.52 191 15.3%
PFE 5.5% 15 Oct 2009 17.66 28 Dec 2009 18.63 75 4.8%
UNH 11.2% 29 Oct 2009 26.47 27 Nov 2009 29.44 30 10.5%
UNH 6.0% 12 Nov 2009 28.76 11 Dec 2009 30.48 30 5.3%
GE 14.8% 27 Nov 2009 15.94 25 Mar 2010 18.3 119 14.1%
BA 8.7% 11 Dec 2009 55.6 12 Jan 2010 60.43 33 8.0%
AA 12.1% 28 Dec 2009 16.1 05 Apr 2011 18.05 464 11.4%
DD 18.3% 12 Jan 2010 34 23 Apr 2010 40.22 102 17.6%
GE 9.5% 25 Mar 2010 18.3 24 Jan 2011 20.04 306 8.8%
AA 14.1% 09 Apr 2010 14.39 07 Jan 2011 16.42 274 13.4%
CVX 4.8% 23 Apr 2010 82.67 09 Dec 2010 86.65 231 4.1%
CAT 12.0% 09 Dec 2010 89.67 07 Feb 2011 100.47 61 11.3%
CSCO 7.6% 23 Dec 2010 19.69 02 Apr 2012 21.19 467 6.9%
BAC -42.7% 07 Jan 2011 14.25 05 Aug 2011 8.17 211 -43.4%
UNH 7.9% 24 Jan 2011 39.81 22 Feb 2011 42.96 30 7.2%
IBM 6.5% 07 Feb 2011 164.82 15 Jul 2011 175.54 159 5.8%
DD -12.5% 22 Feb 2011 54.38 Current 47.58 723 -13.2%
AA -44.0% 05 Apr 2011 18.05 22 Sep 2011 10.11 171 -44.7%
CAT -11.9% 15 Jul 2011 109.36 Current 96.38 580 -12.6%
BAC 16.6% 05 Aug 2011 8.17 19 Mar 2012 9.53 228 15.9%
JPM 6.0% 19 Mar 2012 45 04 Feb 2013 47.68 323 5.2%
BAC 8.1% 02 Apr 2012 9.68 05 Dec 2012 10.46 248 7.3%
HPQ -40.4% 17 Apr 2012 24.72 05 Oct 2012 14.73 172 -41.1%
CSCO 5.5% 05 Dec 2012 19.21 19 Dec 2012 20.27 15 4.8%
HPQ 5.3% 19 Dec 2012 14.38 04 Jan 2013 15.14 17 4.6%
INTC 0.4% 04 Jan 2013 21.16 Current 21.25 41 -0.3%
PG 1.7% 04 Feb 2013 75.25 Current 76.56 10 1.0%



The maximum net profit was BAC @ 90.6% after 32 days
The median net profit was 7.3%


Alicia


Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: klouche Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241714 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 3:06 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 6
At the end the final value without costs was $16,277, representing a gross profit of $5,777 (55%)

First I evaluated this to be correct, given the data on your post.

Then I reduced the trades to annual cash flows, assigning 2013 current positions to 2012.


Year Sum of Bank
2008 -9191.9
2009 186.34
2010 -1362.49
2011 2653.43
2012 13490.81
Grand Total 5776.19


From here using Excel's IRR: 13.1%, a closer approximation to CAGR without trading costs.

Assuming reinvestment of returns (which you did not) gross would have been 85% profit.

If you used Yahoo's rightmost column, adjusted price; you would have dividends included. Only a Yahoo index price excludes dividends.

So this method for this time period did well.

KL

Print the post Back To Top
Author: klouche Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241715 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 3:48 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Regarding Alicia's Dow backtest:

To help visualize what was going on, I calculated the number of positions and days that number was held over the 1816 day period:



Posit Sum of Days
0 2
1 54
2 189
3 377
4 750
5 444
Total 1816


At one point (and for two years) five positions are held, but has declined back to four currently. Knowing this would be important in managing the account.

Anticipating four positions with the caveat that a fifth may be necessary, funded from expected accumulated profits of the four.

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241718 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 5:32 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 3
The sceptics can easily check the dates and prices.

I wish there would be dollar amounts bought and sold for each position. Without that it's difficult to figure out what's going on. The test starts at a very inopportune time. After an initial gaining position, there are four holdings that lost 40% each. One of the positions would have suffered two consecutive 40% losses, which would bring it to 36% of its original value. The overall portfolio would have lost about 50% at that point. It takes a 100% gain to climb out of a 50% loss.

Do you assume that smaller positions were bought, or did you replenish the losses externally? I'm skeptical about reaching a 46.4% profit after that initial loss.

Elan

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241719 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 6:15 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
A few necessary corrections:

1. On June 9, 2009 the symbol TEV should be TRV.

2. All cases where UNH is traded should be changed. UNH was added to the Dow on September 14, 2012. It was not a component of the Dow prior to this.

http://online.wsj.com/article/SB1000087239639044402370457765...

3. TRV was re-added to the Dow on 2009-06-16, so it was not a component on June 9, 2009.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241720 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 6:59 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
klouche (241714)

If you used Yahoo's rightmost column, adjusted price; you would have dividends included. Only a Yahoo index price excludes dividends.

No, I used the "Close" prices column. The "Adj Close" is no help because I may not have been holding the shares within the ex-dividend time frame. Often the dividend arrives after the shares have been sold ( between Ex-D Date and Pay Date), so I need both dates.

My thanks to rharmelink (241687) for providing a link to earnings.com for all the dates needed. I now have to work out the code to extract the details!



elann (241718)

I wish there would be dollar amounts bought and sold for each position. Without that it's difficult to figure out what's going on. The test starts at a very inopportune time. After an initial gaining position, there are four holdings that lost 40% each. One of the positions would have suffered two consecutive 40% losses, which would bring it to 36% of its original value. The overall portfolio would have lost about 50% at that point. It takes a 100% gain to climb out of a 50% loss.

Do you assume that smaller positions were bought, or did you replenish the losses externally? I'm skeptical about reaching a 46.4% profit after that initial loss.


Every purchase was for $10,500/3 = $3,500 (in accordanvce with my rules), and you can calculate the sale$ in each case from the gross profit values. A 40% loss is $1400, so to recover that loss, another $3,500 buy would need to achieve $4,900 - i.e a profit also of 40%. There were 7 stoplosses out of 41 sold holdings, so that is a win ratio of 83%. And there was 1 rebuy - the BAC 5 Aug 2012 in profit at 7.3% net.

Remember that the model treats every transaction independently. I think I have mentioned previously that in practise you would not sell the same company you were about to buy : you would just take profit (eg the BAC 13 Mar 2009 transactions) . This would give a marginal improvement because there would be fewer deals and therefore fewer fees.

No losses were replenished externally. The only funding is the initial $10,500.

The aim of the system has always been to create a long-term investment from a single lump sum.


Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241721 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 7:45 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 6
I see that by Feb 2009 your whole portfolio had enough money for just one position. You bought BAC on 27 Feb 2009. You rolled it over to a new position in BAC on 13 Mar 2009. Then you bought a second position in BAC on 27 Mar 2009. So all your money was in a single stock, BAC, first as one position and then as two concurrent positions in the same stock.

The whole survival and ultimate performance of your strategy depended on the performance of a single stock, on which you conveniently doubled down at just the right time.

I've seen enough. I'm out.

Elan

Print the post Back To Top
Author: rharmelink Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241722 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 8:01 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
No, I used the "Close" prices column.

Be aware of stock splits if you do that...

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241724 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 9:20 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
No, I used the "Close" prices column. The "Adj Close" is no help because I may not have been holding the shares within the ex-dividend time frame. Often the dividend arrives after the shares have been sold ( between Ex-D Date and Pay Date), so I need both dates.

Everything comes out right (and accurate) if you use the "Close" price (adjusted for splits) for generating the signals, and the "Adj Close" price to calculate the returns[*].

If you understand how ex-div works, this statement will be clear to you and you'll understand that it is clearly correct. Conversely, if this is *not* clear to you, then you don't correctly understand how ex-div works.

The only date that matters is the ex-div date. The actual pay-date is of essentially no matter. You'll get the dividend, it's just that the cash may not show up in your account for a few days. FWIW, my broker credits my account on the ex-div date, and then makes corresponding (offsetting) adjustments on the actual pay date.

-----
[*] To be pedantic, the Adj Close assumes that the dividend is immediately re-invested. If your account doesn't reinvest it, but just accumulates the cash, there is a minor by negligible difference in total return.

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241725 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 9:39 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
Elan wrote:
The whole survival and ultimate performance of your strategy depended on the performance of a single stock, on which you conveniently doubled down at just the right time.

In some sense this is true, but actually, the one powerful element of the strategy she is proposing is that it does a great job of selecting the right stock to be invested in at (or near) a market bottom some of the time, as you'll soon see.

I've calculated the Dow 30 for the history she has shared up to the end of 2010. I've actually completed it up to the present, but don't have time to check the results for the last two years.

Any of the lines that are bold show that my pick is not a match to her pick. Sometimes the reason is that the stock was not in the Dow on that date, other times, it may be due to faulty data on her part that did not compensate for splits. Plus, I'm using adjusted data.

So now, instead of holding BAC three times in a row, I end up holding BAC, GM and C.

Alicia     Date     Zeelotes
AA 2/15/2008 INTC
MRK 3/3/2008 MRK
MSFT 3/17/2008 MSFT
DIS 5/13/2008 DIS
AA 11/14/2008 AA
BAC 2/27/2009 BAC
BAC 3/13/2009 GM
BAC 3/27/2009 C
UNH 4/13/2009 C
AXP 4/27/2009 AXP
DIS 5/11/2009 GM
MRK 5/26/2009 MRK
TRV 6/9/2009 MCD
PFE 7/8/2009 PFE
DD 7/22/2009 DD
CAT 8/5/2009 CAT
BAC 8/19/2009 BAC
PG 9/17/2009 PG
GE 10/1/2009 GE
PFE 10/15/2009 PFE
UNH 10/29/2009 XOM
UNH 11/12/2009 PG
GE 11/27/2009 GE
BA 12/11/2009 BA
AA 12/28/2009 AA
DD 1/12/2010 DD
GE 3/25/2010 GE
AA 4/9/2010 AA
CVX 4/23/2010 CVX
CAT 12/9/2010 MRK
CSCO 12/23/2010 CSCO

So for the stocks she had it choosing at the bottom in 2009 the returns were great holding BAC, but when the other two are selected, not so much.

       Entry      Exit     Entry  Exit     ROI    Alicia    ROI
BAC
27-Feb-09 13-Mar-09 3.88 5.68 46.39% BAC 45.10%
GM 13-Mar-09 13-Apr-09 2.72 1.71 -37.13% BAC 90.60%
C 27-Mar-09 27-Apr-09 26.14 30.63 17.18% BAC 20.80%


Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241726 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 10:07 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 3
Remember that the model treats every transaction independently. I think I have mentioned previously that in practise you would not sell the same company you were about to buy

This is a reasonable way to treat this occurrence.

[initial value is] $10,500 ... the final value without costs was $16,277,
Okay.

We really need to see the monthly equity curve. The details on the individual trades is great, but it's impossible to visualize what's happening with the account value from month to month.

The median net profit was 7.3%
That this is reported as an absolute figure rather than a CAGR is cause for concern. It leads me to fear that you are, perhaps unknowingly, doing some sort of illegal math. That's an easy error to make, so you must guard against it.

[initial value is] $10,500 ... the final value $16,277, gross profit of $5,777 (55%)
This is so tremendously large that it boggles the mind. And calls out for close examination.
The DJIA gained 10% in this timeframe. SPY (including dividends) gained 27.7%. SPX (no dividends) gained 14.5%

It is astounding that a trading strategy of DJIA stocks could get 5 times the return. Especially when a half-dozen of the trades had a 40% loss. It takes a 67% gain to recover from a 40% loss, so you'd need half-dozen 67% gains just to get back to even..


Since you didn't show the equity curve, I loaded your trade data into Excel to attempt to create one.

But then I realized that there was an easier way to check the math.
We know that each purchase was $3500, and we have the gain of each trade. From this we can compute the proceeds of each trade.

There were 45 trades. The total cost is then 45 * $3500 or $157,500.
The total proceeds (selling or final price) of the 45 trades totals $162,138.

$167,138 / 157,500 is 3%. Not 55%.

If you take the arithmetic average of the "Net Profit" column, that average is 2.9%. Nowhere near 55%.

Somewhere in the complexity of computing the statistics of your trades, you have made a huge error. You didn't show the math of how you came up with the final value of $16,277, so I am unable to check that math.

Rather than getting 5 times the B&H DJIA return, this strategy UNDERPERFORMED by 3 times.


Of course, I might be wrong. I may mave made a mistake somewhere. If anyone wants to check my math -- or my logic -- please do so.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241728 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 11:42 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
I've now completed the check of the actual stocks selected at each date that Alicia has provided. This says nothing about whether these dates are correctly chosen or not. Of the 45 dates 24.44% of the stocks selected were selected in error per the strategy she has outlined.

As I've already stated, sometimes this was due to survivorship bias. She simply did not use the historical Dow list for each date. At other times, the difference may be due to using adjusted data, but this is probably rarely the case.

Alicia     Date     Zeelotes  Correct?
AA 2/15/2008 INTC
Error
MRK 3/3/2008 MRK
MSFT 3/17/2008 MSFT
DIS 5/13/2008 DIS
AA 11/14/2008 AA
BAC 2/27/2009 BAC
BAC 3/13/2009 GM Error
BAC 3/27/2009 C Error
UNH 4/13/2009 C Error
AXP 4/27/2009 AXP
DIS 5/11/2009 GM Error
MRK 5/26/2009 MRK
TRV 6/9/2009 MCD Error
PFE 7/8/2009 PFE
DD 7/22/2009 DD
CAT 8/5/2009 CAT
BAC 8/19/2009 BAC
PG 9/17/2009 PG
GE 10/1/2009 GE
PFE 10/15/2009 PFE
UNH 10/29/2009 XOM Error
UNH 11/12/2009 PG Error
GE 11/27/2009 GE
BA 12/11/2009 BA
AA 12/28/2009 AA
DD 1/12/2010 DD
GE 3/25/2010 GE
AA 4/9/2010 AA
CVX 4/23/2010 CVX
CAT 12/9/2010 MRK Error
CSCO 12/23/2010 CSCO
BAC 1/7/2011 BAC
UNH 1/24/2011 HPQ Error
IBM 2/7/2011 IBM
DD 2/22/2011 DD
AA 4/5/2011 AA
CAT 7/15/2011 CAT
BAC 8/5/2011 JPM Error
JPM 3/19/2012 JPM
BAC 4/2/2012 BAC
HPQ 4/17/2012 HPQ
CSCO 12/5/2012 CSCO
HPQ 12/19/2012 HPQ
INTC 1/4/2013 INTC
PG 2/4/2013 PG

Error 11
% Error 24.44%


Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: FlyingCircus Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241729 of 253091
Subject: Re: Does this stuff work ? Date: 2/19/2013 11:59 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Thank you for posting detail trades on that trending system.

You'll no doubt notice that that forced buy system makes 3 purchases at nearly the absolute bottom of the Meltdown, and 2 more within a month. These cause major outlier results.

I wanted to see how it did exclusive of the 08 melt, forced buys and 56% market bounce from March to September 2009. So just for shiggles, I ran your trades above starting with GE 10/1 2009 and your allocation system (starting $x/3).

With positions rolling over into new buys, etc. I get a total return of 31.9% since 10/1/09 which isn't bad.

The Dow since 10/1/09 is up 48%.

FC

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241731 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 1:42 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 8
There were 45 trades. The total cost is then 45 * $3500 or $157,500.
The total proceeds (selling or final price) of the 45 trades totals $162,138.

$167,138 / 157,500 is 3%. Not 55%.

If you take the arithmetic average of the "Net Profit" column, that average is 2.9%. Nowhere near 55%.


You are of course grossly mistaken. This portfolio did not start out with $157500 and end with $167138 five years later for a total return of 3%, as your erroneous math implies.

It held 45 positions over five years, which is an average of 9 positions each year. If we assume 3 concurrent positions (the actual average is a bit higher) then the whole portfolio turned over 3 times a year on average. And it made an average 3% gain on each turnover. That works out to a compounded return of 1.03^3=1.0927, or 9.27% per year. Over five years that's 1.0927^5=1.558, or a total return of 55.8%.

Presto! Bingo!

Elan

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241733 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 8:52 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
What a pain - I see my response posting missed a crucial formatting character - and this website has no post-publishing editing facility.
Please disregard previous post or maybe the moderator will delete it?

Hey Guys,

I see some of you now understand the system.
I have looked through yesterday's posts and my replies are as follows:


Rayvt (241724)
To be pedantic, the Adj Close assumes that the dividend is immediately re-invested. If your account doesn't reinvest it, but just accumulates the cash, there is a minor by negligible difference in total return.
Also Klouche (241714)
If you used Yahoo's rightmost column, adjusted price; you would have dividends included.

Still on the dividends issue, presumably I would need to use the Yahoo "close" price for buys and the "adj price" for sales (it would not be correct to use a dividend-adjusted price as a buying price). My system of loading XD and Pay dates avoids this issue by crediting the cash-at-bank only when a dividend is actually received. That extra cash then adds to the funds for the next target buy which could be the next day. As you see, the timing of the dividend payment is important to the scheme.


Zeelotes (241719)
A few necessary corrections:

Thanks for that. It's a fine point but I do not believe it significantly discredits the result.

elann (241721)

I see that by Feb 2009 your whole portfolio had enough money for just one position. You bought BAC on 27 Feb 2009. You rolled it over to a new position in BAC on 13 Mar 2009. Then you bought a second position in BAC on 27 Mar 2009. So all your money was in a single stock, BAC, first as one position and then as two concurrent positions in the same stock.

The whole survival and ultimate performance of your strategy depended on the performance of a single stock, on which you conveniently doubled down at just the right time.


Point taken.
So, I have re-run on the basis of 5 startups instead of 3.
Of course, after a similar start, as you would expect, the target buys taken up are completely different from previously.
The final result still shows 54% gross, but as there were now 68 buys the costs are higher and the net profit is down a bit at 41.3%. There were 9 stoploss sales and the win ratio for holdings sold is higher at 85%.

Still a good result though.



rharmelink (241722)
No, I used the "Close" prices column.

Be aware of stock splits if you do that...


Reasonable comment - thanks.
The only one I came across was 2for1 in KO which was advertised as 17 Jul 2012 but actioned in Yahoo 13 Aug 2012. As it happens KO was not selected as a target buy.

elann (241725)
Any of the lines that are bold show that my pick is not a match to her pick. Sometimes the reason is that the stock was not in the Dow on that date, other times, it may be due to faulty data on her part that did not compensate for splits. Plus, I'm using adjusted data.

I have simply used the raw Dow data from Yahoo (there was only he KO split to adjust). If that has errors there is nothing I can do about it.
Obviously if you are using adjusted data you are effectively using a completely different dataset and would expect to get a different result.


Rayvt (241726)

We really need to see the monthly equity curve. The details on the individual trades is great, but it's impossible to visualize what's happening with the account value from month to month.


I am sure you could work it out for yourself, but as I have already done so I will save you the trouble.
Of course you will have to trust me though.
OK, Here you are:

	        Bank	Holdings Overall
31-Jan-08 10,500 0 10,500
29-Feb-08 7,000 3,639 10,639
31-Mar-08 0 10,058 10,058
30-Apr-08 0 9,957 9,957
30-May-08 627 10,023 10,650
30-Jun-08 627 9,577 10,204
31-Jul-08 627 8,889 9,516
29-Aug-08 627 9,507 10,134
30-Sep-08 627 8,937 9,564
31-Oct-08 2,709 5,403 8,112
28-Nov-08 1,264 5,975 7,239
31-Dec-08 1,264 6,040 7,304
30-Jan-09 1,264 4,630 5,895
27-Feb-09 1,957 3,500 5,457
31-Mar-09 61 7,396 7,457
30-Apr-09 4,011 7,021 11,032
29-May-09 916 10,655 11,570
30-Jun-09 1,306 10,251 11,557
31-Jul-09 1,803 10,783 12,586
31-Aug-09 2,558 10,764 13,322
30-Sep-09 3,102 11,076 14,178
30-Oct-09 260 13,587 13,847
30-Nov-09 1,016 14,119 15,135
31-Dec-09 1,418 13,550 14,967
29-Jan-10 1,722 13,179 14,901
26-Feb-10 1,722 13,409 15,130
31-Mar-10 2,240 14,399 16,639
30-Apr-10 3,439 13,241 16,680
28-May-10 3,439 11,616 15,055
30-Jun-10 3,439 10,265 13,704
30-Jul-10 3,439 11,455 14,894
31-Aug-10 3,439 10,613 14,052
30-Sep-10 3,439 12,117 15,556
29-Oct-10 3,439 12,613 16,052
30-Nov-10 3,439 12,504 15,942
31-Dec-10 107 17,839 17,946
31-Jan-11 934 18,129 19,063
28-Feb-11 1,632 17,441 19,074
31-Mar-11 1,632 17,162 18,795
29-Apr-11 2,056 16,704 18,761
31-May-11 2,056 16,150 18,206
30-Jun-11 2,056 15,664 17,720
29-Jul-11 2,284 14,551 16,835
31-Aug-11 791 14,787 15,577
30-Sep-11 2,751 10,313 13,064
31-Oct-11 2,751 12,337 15,088
30-Nov-11 2,751 11,848 14,599
30-Dec-11 2,751 11,442 14,193
31-Jan-12 2,751 13,315 16,066
29-Feb-12 2,751 13,876 16,627
30-Mar-12 3,334 14,150 17,483
30-Apr-12 100 16,511 16,611
31-May-12 100 14,357 14,458
29-Jun-12 100 14,556 14,657
31-Jul-12 100 13,930 14,031
31-Aug-12 100 14,101 14,201
28-Sep-12 100 14,746 14,846
31-Oct-12 2,186 12,191 14,377
30-Nov-12 2,186 12,265 14,451
31-Dec-12 2,661 12,651 15,312
31-Jan-13 2,846 13,342 16,189
13-Feb-13 3,054 13,223 16,277




Zeelotes (241728)
Of the 45 dates 24.44% of the stocks selected were selected in error per the strategy she has outlined.

What errors? I used only the Dow data from Yahoo.
Please explain.


Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: klouche Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241734 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 9:18 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Still on the dividends issue, presumably I would need to use the Yahoo "close" price for buys and the "adj price" for sales (it would not be correct to use a dividend-adjusted price as a buying price). My system of loading XD and Pay dates avoids this issue by crediting the cash-at-bank only when a dividend is actually received. That extra cash then adds to the funds for the next target buy which could be the next day. As you see, the timing of the dividend payment is important to the scheme

Use adjusted price on buy and sell, the dividend is posted to your account later if the ex-div date occurs while holding and is posted even if you sell the stock after the ex-div date but before posting. There's no loyalty check. You can use either close or adjusted close for the set-up.

I'm evaluating this system purely on the money at risk, to compare with our screens; money in the bank is not relevent.

The bar is low here, performance wise. Here's how some of our old screens performed during this era, if they had a large cap filter:


Screen 08-13
CAPRS -11
CAPLOWEG -9
HBSP 0
Keystone 2
Blitz -4
HIYIELD 2

Average -3


However the surviviorship bias error is serious. You have to know the components of the DOW in 2008 and shift to new components as they are defined in the following years. If you are holding a component that the DOW drops, it's up to you if you sell immediately and swap or hold until your exit criteria is met.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241735 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 9:21 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
$167,138 / 157,500 is 3%.

Dummy me! This is the wrong formula.

The right formula would be final_value vs. starting_value.

Which is (167,138 - 157,500) / 10,500.

Which is 9638/10500 or -8.2%.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241736 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 9:35 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Rayvt (241735)
The right formula would be final_value vs. starting_value.

Correct.

But your figures have no meaning in the context of my system.

My final_value = 16,227
My starting_value = 10,500

16227/10500 = 1.5455 ie 54.5% gross profit

Alicia

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241739 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 10:05 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Still on the dividends issue, presumably I would need to use the Yahoo "close" price for buys and the "adj price" for sales (it would not be correct to use a dividend-adjusted price as a buying price).

It depends how you track the data in your backtest. You are right, of course, in saying that the buy price would be the "close" price. The sell price is also the close price, and the total return is the difference between sell and buy, PLUS any dividend earned.

The other way to compute it is to note in passing (and then subsequently ignore) that the buy was for X number of shares, where X = $3500 (approx) divided by "close" price.
The gain on that position is the adjclose price as of the sell date divided by the adjclose price as of the buy date.

My system of loading XD and Pay dates avoids this issue by crediting the cash-at-bank only when a dividend is actually received. That extra cash then adds to the funds for the next target buy which could be the next day. As you see, the timing of the dividend payment is important to the scheme.
In the big scheme of things, the few days difference between the pay date and the ex-div date is negligible. The standard way that this is handled by EVERYONE is to treat the dividend as if it was credited to you on the ex-div date.
It is important to not get buried in unimportant details, because that leads to complexity that can lead to errors. Look up "false precision".

So let's look at a random DJIA stock and see what's what.
CVX has 3.1% yield. It went ex-div on 2/13/13 and paid 0.90 dividend.
CVX is about 115, so a $3500 position would hold 30 shares and the dividend would be $27. Finding the actual pay date is difficult, but I finally found it. March 11.
So the question to ask is: "For an amount of $27, will a timing difference of 3 weeks have a significant effect on the overall portfolio?" I would submit that it is negligible.
Most of the time, thet $27 won't even buy one additional share.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241740 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 10:12 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
klouche (241734)
However the surviviorship bias error is serious. You have to know the components of the DOW in 2008 and shift to new components as they are defined in the following years. If you are holding a component that the DOW drops, it's up to you if you sell immediately and swap or hold until your exit criteria is met.

In my UK dealings I buy only those in the current FTSE100 list at the date of target buy detection, but I continue to hold them when they drop out until they reach the primary profit position for selling to fund a new buy.

If you can provide me a link to an historical DJ components list I could do the same.


Alicia

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241741 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 10:19 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Rayvt (241739)
So the question to ask is: "For an amount of $27, will a timing difference of 3 weeks have a significant effect on the overall portfolio?" I would submit that it is negligible.
Most of the time, thet $27 won't even buy one additional share.


At a possible buy every 10 days we are dealing here at with at least 130 possible target buys. At $27 per buy that could be at least £3,500. That's significant.

Alicia

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241742 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 10:32 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 10
Alicia asked:
What errors? I used only the Dow data from Yahoo.
Please explain.


If you can provide me a link to an historical DJ components list I could do the same.

From your comments it seems you do not fully understand the importance of point-in-time historical component lists.

In order to do a backtest of any trading system you must have the actual list on the historical date you are considering for entry into a new position. This list is easy to obtain for the Dow, but very difficult, if not impossible for most people, for all other indexes. I would guess that this is also true for the FTSE100 list. If you have a link to the historical additions and deletions from this list please provide it.

Here is the fruit of a simple search on Google:

"Dow Historical Components"

http://en.wikipedia.org/wiki/Historical_components_of_the_Do...

It is probably not 100% accurate as it relates to symbol changes and mergers/acquisitions, but will be much, much more accurate than using today's list to do a backtest from 2008.

So the errors I pointed out above, are those cases where you were using today's list on a date in the past, when that particular stock was NOT in the Dow 30. There would have been no way for you to know that that stock would later be added. In other words, on that date, you never would have even considered purchasing that stock.

For example, on 3/13/2009 you purchased BAC which had a 90.60% gain. In fact, on that date GM beat out BAC using your TREND formula so you would have bought that and not BAC. The result of holding that for the same period would have resulted in a loss of -37.13%. I realize that the actual holding of GM would not have been for the same time period, but rather, would have actually resulted in another -40% stoploss.

Do you now see why knowing the historical Dow list can make or break any trading system like you are using?

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241743 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 10:53 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
We really need to see the monthly equity curve. The details on the individual trades is great, but it's impossible to visualize what's happening with the account value from month to month.


I am sure you could work it out for yourself, but as I have already done so I will save you the trouble.
Of course you will have to trust me though.


Thanks. Taking these values as correct, I loaded them into a spreadsheet to graph the equity curve. It can be viewed/downloaded here:
https://docs.google.com/file/d/0B1yWRtMroxvgaXhiZEMtRVd1bVk/...

I also plotted the S&P500 for the same dates with the same scale, so I could visualize how sell-to-buy behaved in comparison to the S&P.

Here's what troubled me: The plots were identical from 1/08 thru 2/09, at which time STB shot skyward while S&P just muddled along. But...the two lines seemed to have similar shape, with STB much higher that S&P.

So I created another chart with a starting date of 4/09, which was 2 months after STB took that large jump. Lo and behold, STB and S&P tracked quite closely, sometimes one ahead, sometimes the other.

The entire outperformance of STB was due to one 2-month period, where STB fortuitously picked BAC.

There is a word for this, and that word is "LUCK".

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241744 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 11:33 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
Rayvt wrote:
The entire outperformance of STB was due to one 2-month period, where STB fortuitously picked BAC.

Here is the sort on 3/13/2009 when BAC was chosen. In fact, GM is the one that would have been selected on that date. It then would have resulted in hitting the stop-loss with a forty percent loss.

                  3/13/2009              Trend
GM General Motors Corp 2.15%
BAC Bank Of America Corp 1.22%
CAT Caterpillar Inc 0.90%
PFE Pfizer Inc 0.80%
DIS Disney Walt Co 0.28%
WMT Wal Mart Stores Inc 0.25%
MMM 3m Co 0.17%
T A T & T Inc 0.14%
KFT Kraft Foods Inc -0.07%
PG Procter & Gamble Co -0.09%
HD Home Depot Inc -0.16%
INTC Intel Corp -0.22%
VZ Verizon Communications Inc -0.32%
MCD Mcdonalds Corp -0.34%
KO Coca Cola Co -0.43%
AA Alcoa Inc -0.53%
GE General Electric Co -0.57%
UTX United Technologies Corp -0.63%
XOM Exxon Mobil Corp -0.68%
IBM International Business Machs Cor -0.73%
CVX Chevron Corp New -0.83%
MSFT Microsoft Corp -1.00%
DD Du Pont E I De Nemours & Co -1.04%
JNJ Johnson & Johnson -1.18%
BA Boeing Co -1.26%
AXP American Express Co -1.30%
JPM Jpmorgan Chase & Co -1.38%
HPQ Hewlett Packard Co -1.38%
MRK Merck & Co Inc -2.00%
C Citigroup Inc -10.70%


Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241745 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 11:38 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
The right formula would be final_value vs. starting_value.

Which is (167,138 - 157,500) / 10,500.
Which is 9638/10500 or -8.2%.

Wrong wrong wrong. ::whacks own head::

You netted $9638. You collected 167,138 but only paid 157,500. That means you have $9638 more than you started with.

Now you see why I need to visualize things in simple terms. ;-)

That'll teach me to try to do math right after I get up in the morning. ...sigh.....

But that leads to another problem. Alicia says her final value was $16,277 for a gain of $5777, and my simplistic naive calculation using her list of trades (using a constant $3500 for opening each position) comes up with a gain of $9638. That's not only not in the same ballpark, it's not even in the same county.

Print the post Back To Top
Author: hk2 Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241746 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 11:50 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
and my simplistic naive calculation using her list of trades (using a constant $3500 for opening each position) comes up with a gain of $9638


Per post 241713, from 31 Mar 08 - 31 Mar 09 there were 12 trades where the balance was below the starting amount. ($10500)

Alicia previously stated there was no external funding, so I'm failing to see how every buy is $3500.



Jim

Print the post Back To Top
Author: warrl Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241747 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 11:55 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Still on the dividends issue, presumably I would need to use the Yahoo "close" price for buys and the "adj price" for sales (it would not be correct to use a dividend-adjusted price as a buying price).

Sadly, it's not that simple.

Yahoo's "adj price" is always the same as the closing price for the last day that Yahoo has historical prices available for, at the moment you get the dataset (which usually means the prior trading day).

If there is a dividend after your sell date but before you get the data, then you can't just use the adjusted price on the sell date as it reflects dividends you did not and will not collect. Later splits would make a similar but much more obvious mess.

Here's how you calculate an adjustment factor:

(sell date adjusted closing price) / (sell date closing price)
--------------------------------------------------------------
(buy date adjusted closing price) / (buy date closing price)


To calculate adjusted closing price on sell date relative to closing price on buy date, multiply the closing price on sell date by the adjustment factor.

Or, to calculate adjusted closing price on buy date relative to closing price on sell date, divide the closing price on buy date by the adjustment factor.

DON'T DO BOTH; the result would be meaningless.

(Yes, I ran this through a spreadsheet and compared the result against calculating the number of shares purchased with reinvested dividends and tracing the value of the position. I'm quite confident it's right. However, I'm also quite aware that I'm capable of error.)

Oh, and this calculation assumes (a) there is no significant change in price, other than those caused by splits, between the ex-dividend date and the date the dividend is paid, and (b) you immediately reinvest the dividend in the same stock.

If you instead keep dividends as cash or invest them in something else, you need to go through the history day-by-day watching for changes in the money amount gap between closing price and adjusted close. Small changes represent dividends paid, and the dividend per share is the amount of the change. Large changes represent stock splits. Hopefully you don't get handed both on the same day on one stock...

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: DoesMIWork One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241748 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 12:23 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Amazing insight!

Just looking at all the things that go wrong is mind boggling.

This is why I favor deep screens, like 15 picks/screen. Any change in one of the picks can drastically shift the landscape. A three deep screen with 65% CAGR can easily be a -10% CAGR with a single pick blowing up.

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241749 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 12:27 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 12
Elan wrote:
It held 45 positions over five years, which is an average of 9 positions each year. If we assume 3 concurrent positions (the actual average is a bit higher) then the whole portfolio turned over 3 times a year on average. And it made an average 3% gain on each turnover. That works out to a compounded return of 1.03^3=1.0927, or 9.27% per year. Over five years that's 1.0927^5=1.558, or a total return of 55.8%.

Using your approach here are the results for this system if BAC was actually chosen on March 2009, and if GM was the trade which is what the system would have selected.

          Average  Per Year  Five Years
With GM
0.04% 0.13% 0.64%
With BAC 2.94% 9.10% 54.54%


One trade turned a profitable system into rubbish. Ah, the importance of integrity in data and implementation!

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241750 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 1:50 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Zeelotes (241742)
I would guess that this is also true for the FTSE100 list. If you have a link to the historical additions and deletions from this list please provide it..

Also from a simple Google search:

https://www.ftse.com/Media_Centre/Downloads/FTSE_100_Constit...

And thanks for your link to Dow constituents history.



So the errors I pointed out above, are those cases where you were using today's list on a date in the past, when that particular stock was NOT in the Dow 30

I am not sure that is really an issue for me because I have simply been trying to get you to try my system on a dataset of companies. I only chose the Dow 30 because they seemed equivalent to the FTSE 100 constituents in terms of highly liquid stocks which would be easily traded and with which you would be familiar. You might as well use any database where the stocks have high volatility, lending themselves to trend analysis. My system is really just a data manipulation tool, but I am trying not to get acceptance for the stock selection method, rather I am looking for the STB to be validated as a way of selling stocks effectively.

You could use a longer trend formula on the Dow data, say 30 day instead of 20 day.
TrendA=TREND(A1:A30,,{31})/TREND(A1:A30,,{30})-1
TrendB=TREND(A30:A59,,{31})/TREND(A30:A59,,{30})-1

On the same Dow dataset as before, this results in a final value of $19,459 from the starting $10,500, a gross profit of 85% (net 74.7%) with a total of 56 stocks being bought (win ratio on sold holdings 88%) with 5 current.

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241751 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 2:28 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Alicia -- Did you use this FTSE100 list to build your backtest? And do you have historical data on each of these companies for the full history of your backtest?

My system is really just a data manipulation tool, but I am trying not to get acceptance for the stock selection method, rather I am looking for the STB to be validated as a way of selling stocks effectively.

It is very difficult to validate the method for selling stocks. I'm still unclear of the parameters for doing so.

If you are holding three stocks, and one of them hits a return of 5% sell it? It seems that all of the trades would then be at what we call a Profit Latch of 5%. Clearly that is not what you are doing.

Please state the rules and give concrete examples.

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241752 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 2:38 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
So the errors I pointed out above, are those cases where you were using today's list on a date in the past, when that particular stock was NOT in the Dow 30

I am not sure that is really an issue for me because I have simply been trying to get you to try my system on a dataset of companies.
...
On the same Dow dataset as before


You'd better face reality, because if you change your Dow list from a fantasy list to the real components the outcome changes from gold to dust.

Don't ignore that. Don't backpedal by claiming after the fact that the Dow experiment was just an example and you can ignore it and pretend that you have a working system. You don't. So far it has failed in half your backtests.

Elan

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241753 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 2:41 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Per post 241713, from 31 Mar 08 - 31 Mar 09 there were 12 trades where the balance was below the starting amount. ($10500)

Alicia previously stated there was no external funding, so I'm failing to see how every buy is $3500.


If the account balance drops below 10500 the number of stocks held is shrunk to 2 or even to 1. If the account balance shrinks below 3500 the system collapses.

Elan

Print the post Back To Top
Author: warrl Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241754 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 3:17 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
If the account balance shrinks below 3500 the system collapses.

Technically, not necessarily.

As I understand it there are two sell signals:

a) a 40% decline

b) something is up 5% from its purchase price AND you have an active buy signal for something else (although I'm not clear on all aspects of the definition of an active buy signal).

Since you can't ever get a sell signal of the second type unless at least one position is UP from its initial value of 3500, the system won't collapse because of that.

So for the system to definitively collapse, all extant positions have to decline by 40% and deliver a cash-out total value under 3500. Absent a significant gap, two such positions will cash out for more than 3500 so it would have to be when you're already beaten down to only one position.

That CAN happen.

However, if your last position declines merely 39% (from 3500 to 2135) and then stays absolutely flat until your great-grandchild retires, technically the system is still functioning. Who knows, the one stock COULD pop up 72% to enable a profitable sell signal...

Print the post Back To Top
Author: klouche Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241755 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 3:49 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Technically, not necessarily.

As I understand it there are two sell signals:

a) a 40% decline

b) something is up 5% from its purchase price AND you have an active buy signal for something else (although I'm not clear on all aspects of the definition of an active buy signal).

Since you can't ever get a sell signal of the second type unless at least one position is UP from its initial value of 3500, the system won't collapse because of that.

So for the system to definitively collapse, all extant positions have to decline by 40% and deliver a cash-out total value under 3500. Absent a significant gap, two such positions will cash out for more than 3500 so it would have to be when you're already beaten down to only one position.

That CAN happen.

However, if your last position declines merely 39% (from 3500 to 2135) and then stays absolutely flat until your great-grandchild retires, technically the system is still functioning. Who knows, the one stock COULD pop up 72% to enable a profitable sell signal...


To end this endless thread, we have to separate the account management aspect of the system and isolate the efficacy of the set up.

To do that a simple table:

                              Forward Return Days
Signal Level Bucket 21 42 63 181 252
x 1+
y 0-1
z less than zero


Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241756 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 6:16 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 6
I only chose the Dow 30 because they seemed equivalent to the FTSE 100 constituents in terms of highly liquid stocks which would be easily traded and with which you would be familiar. You might as well use any database where the stocks have high volatility, lending themselves to trend analysis. My system is really just a data manipulation tool, but I am trying not to get acceptance for the stock selection method,

Alicia, you are still not getting it.

What people have been showing is that it is HARD to get accurate historical data, and EASY to get bad data.
It is EASY to make methodological errors in the backtests, such as using a stock that is in your universe (DOW, FTSE100, or whatever) NOW that wasn't in the universe at the point-in-time that your backtest made a buy decision.

You seem to not recognise that these types of problems exist. When people point them out and/or ask you how you addressed it ---- you completely ignore them


I am looking for the STB to be validated as a way of selling stocks effectively.
Two responses here:
1) That can't be done. It is not possible to separate the selling rules from the buying rules.
2) Your selling rules fail. Period. You sell a stock when it shows a 5% profit, or it shows a 40% loss. It is without question that the general rule for success in selling is to keep the stocks that are going up and sell the stocks that are going down. It is without question that when you *must* sell a holding, the best strategy is to sell your worst performing stock.

Yet your scheme does the exact opposite. You claim that doing the converse is the best strategy, and give NO reason or rationale why.

At this point you are becoming just a troll. This is the third thread where you've touted your sell-to-buy strategy (2 Brit forums and this forum). I will grant that you've posted trade data -- which is more than the typical troll does -- but you have never given any cogent reasoning or rationale for this strategy -- and you have been given plenty of opportunity to. People on all 3 forums discussed details, checked math, and asked questions to clarify details.

You never engage in debate or discussion. You ultimate fallback position is asking people to "look at their own portfolio and see how sell-to-buy would have improved their returns."

I am looking for the STB to be validated as a way of selling stocks effectively.
This is good. You are asking people -- people who presumably are experienced investors -- if your proposed stratgy is valid.
The unanimous opinion is "No."

So why are you still asking? Why are you still proposing that people try your strategy?

For your sake, and for the sake of your grandchildren that you are investing for, STOP using this strategy. It is not going to outperform a buy-and-hold strategy. It is going to lose your money.

And stop being a troll.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: musselmant Three stars, 500 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241757 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 6:23 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
since 2/93 BRK.a is up 1,173% v. 253% for the S&P

We need to find something better or else just let Warren do the buying for us.

Print the post Back To Top
Author: KBGlenn Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241760 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 7:08 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
I had to post just to be part of his epic thread. I'm happy it eventually got to data...

Print the post Back To Top
Author: alevine Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241761 of 253091
Subject: Re: Does this stuff work ? Date: 2/20/2013 7:50 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 40
This thread is important not because it finds or refutes another stock picking/selling system.

It is important because it shows just how crucial having accurate underlying data really is to anything we try and do here.

I'm not an active poster these days on the MI Board, but I follow it daily (at least!). But I was one of the folks who was intimately involved in constructing the VL/CRSP database some 15 years ago.

I think it's easy to underestimate the time and effort that goes into making sure we have good data on which to perform our backtests....

.....but without it, we've got nothing!


Alan

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241767 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 9:01 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 3
Zeelotes (241751)
It is very difficult to validate the method for selling stocks. I'm still unclear of the parameters for doing so.

If you are holding three stocks, and one of them hits a return of 5% sell it? It seems that all of the trades would then be at what we call a Profit Latch of 5%. Clearly that is not what you are doing.

Please state the rules and give concrete examples.



The 20-day trend function for finding target buys caused lots of criticism, much about the dataset rather than the process, and I have done my best not to respond to all the naysayers. However, there have also been some helpful comments. Incidentally, on the BAC issue my results were correct but in retrospect I have realised that I was using a slightly different 20-day trend function than I had posted.

OK, I will put it all together with another example, but I can only give the example using the historical Yahoo DOW data close prices for current constituents (year to Feb 2013) since I don't have any more accurate data - i.e the same data I have used before in this thread.

For this example lets look at a 30-day trend function :

TrendA=TREND(A1:A30,,{31})/TREND(A1:A30,,{30})-1
TrendB=TREND(A31:A60,,{31})/TREND(A31:A60,,{30})-1

Look across each day of the dataset for the highest value of TrendB-TrendA to find the target buy.
Repeat every 10 days for the next target buy.


My Sell-to-Buy rules are as follows (and I have purposely omitted the crash rule since I did not use that in posting previous results):

1. Divide an initial investment sum by 3 to give a lot size for all subsequent purchases.
2. Buy your first three holdings.
3. To buy your 4th holding, only do so if you best holding is showing at least 5% profit.
4. Bank the profit and continue to accumulate profits each time you sell.
5. Buy next holding from banked profits when possible
6. If bank is insufficient and no holding is in profit when you want to buy, do nothing.
7. Apply a 40% stoploss to each holding.
8. Track your stoploss companies (including those you did not buy) and rebuy if they fall below 50% of original price.


the following is a schedule of the results :


EPIC Gain Bought @ Sold @ Held Net Profit
HD 7.2% 15 Feb 2008 27.52 01 Apr 2008 29.49 47 5.9%
AA 9.9% 03 Mar 2008 38.32 13 May 2008 42.12 72 9.2%
INTC 8.5% 17 Mar 2008 20.85 29 Apr 2008 22.62 44 7.8%
MSFT -40.6% 01 Apr 2008 29.5 20 Nov 2008 17.53 234 -41.3%
UNH -42.5% 29 Apr 2008 33.4 08 Oct 2008 19.2 163 -43.2%
UNH 5.2% 13 May 2008 32.55 28 May 2008 34.23 16 4.4%
AA -40.1% 28 May 2008 41.57 25 Sep 2008 24.92 121 -40.8%
JPM 4.9% 17 Oct 2008 39.33 31 Oct 2008 41.25 15 4.2%
UNH 13.9% 31 Oct 2008 23.73 30 Dec 2008 27.04 61 13.2%
UNH 5.2% 20 Nov 2008 16.3 21 Nov 2008 17.15 2 4.5%
JPM 15.0% 21 Nov 2008 22.72 01 Dec 2008 26.12 11 14.3%
AA 6.4% 01 Dec 2008 9.31 15 Dec 2008 9.91 15 5.7%
AA -41.4% 15 Dec 2008 9.91 23 Feb 2009 5.81 71 -42.1%
CAT -42.5% 30 Dec 2008 43.66 23 Feb 2009 25.12 56 -43.2%
JPM 19.9% 27 Feb 2009 22.85 27 Mar 2009 27.4 29 19.2%
BAC 50.1% 27 Mar 2009 7.34 13 Apr 2009 11.02 18 49.4%
BAC 17.4% 13 Apr 2009 11.02 11 May 2009 12.94 29 16.7%
AXP 11.1% 27 Apr 2009 24.23 09 Jun 2009 26.93 44 10.4%
AXP 10.4% 11 May 2009 26.04 22 Jul 2009 28.76 73 9.7%
MRK 9.0% 09 Jun 2009 25.72 08 Jul 2009 28.04 30 8.3%
VZ 9.1% 08 Jul 2009 28.65 05 Aug 2009 31.27 29 8.4%
AXP 10.2% 22 Jul 2009 28.76 19 Aug 2009 31.7 29 9.5%
CAT 17.4% 05 Aug 2009 46.64 15 Oct 2009 54.77 72 16.7%
CAT 19.3% 19 Aug 2009 45.17 17 Sep 2009 53.89 30 18.6%
BA 7.7% 02 Sep 2009 48.4 01 Oct 2009 52.11 30 7.0%
BA 14.3% 17 Sep 2009 52.88 12 Jan 2010 60.43 118 13.6%
PG 5.2% 01 Oct 2009 56.62 29 Oct 2009 59.54 29 4.4%
PG 7.1% 15 Oct 2009 57.24 12 Nov 2009 61.3 29 6.4%
XOM 13.5% 29 Oct 2009 73.96 07 Feb 2011 83.93 467 12.8%
UNH 6.0% 12 Nov 2009 28.76 11 Dec 2009 30.48 30 5.3%
BA 11.4% 11 Dec 2009 55.6 27 Jan 2010 61.93 48 10.7%
AA 6.6% 12 Jan 2010 15.52 22 Feb 2011 16.54 407 5.9%
JPM 9.8% 27 Jan 2010 39.33 11 Mar 2010 43.18 44 9.1%
DIS 7.1% 11 Mar 2010 33.81 09 Apr 2010 36.22 30 6.4%
AA 8.7% 25 Mar 2010 14.11 23 Dec 2010 15.34 274 8.0%
AA 14.1% 09 Apr 2010 14.39 07 Jan 2011 16.42 274 13.4%
BAC 6.6% 23 Dec 2010 13.06 24 Jan 2011 13.92 33 5.9%
CSCO 0.8% 07 Jan 2011 20.97 Current 21.14 769 0.1%
CSCO -0.1% 24 Jan 2011 21.17 Current 21.14 752 -0.9%
UNH 4.8% 07 Feb 2011 41.81 08 Mar 2011 43.83 30 4.1%
TEV 5.0% 22 Feb 2011 60.53 04 May 2011 63.54 72 4.3%
KO 5.2% 08 Mar 2011 32.82 23 Aug 2011 34.53 169 4.5%
INTC 7.9% 04 May 2011 23.5 06 Dec 2011 25.35 217 7.2%
BAC 23.2% 23 Aug 2011 6.3 26 Aug 2011 7.76 4 22.5%
PG 5.8% 26 Aug 2011 62.57 20 Jan 2012 66.23 148 5.1%
UNH 8.9% 06 Dec 2011 48.29 05 Jan 2012 52.59 31 8.2%
UNH 9.8% 20 Dec 2011 49.82 17 Feb 2012 54.68 60 9.0%
GE 7.9% 05 Jan 2012 18.55 02 Apr 2012 20.02 89 7.2%
BAC 10.9% 20 Jan 2012 7.07 03 Feb 2012 7.84 15 10.2%
BAC 21.6% 03 Feb 2012 7.84 19 Mar 2012 9.53 46 20.8%
HPQ -40.3% 17 Feb 2012 29.59 01 Aug 2012 17.66 167 -41.0%
PG 12.0% 19 Mar 2012 67.21 04 Feb 2013 75.25 323 11.2%
JPM 5.5% 02 Apr 2012 46.13 Current 48.68 318 4.8%
PG 4.8% 09 Aug 2012 66.73 18 Jan 2013 69.94 163 4.1%
HPQ -0.6% 18 Jan 2013 17.11 Current 17.01 27 -1.3%
HD 1.6% 04 Feb 2013 66.36 Current 67.45 10 0.9%



and this shows the monthly data:


Bank Holdings # held Overall Gain
10,500 0 0 10,500 0%
29-Feb-08 7,000 3,377 1 10,377 -1%
31-Mar-08 0 10,406 3 10,406 -1%
30-Apr-08 548 9,980 3 10,527 0%
30-May-08 1,075 10,362 3 11,438 9%
30-Jun-08 1,075 9,014 3 10,089 -4%
31-Jul-08 1,075 8,836 3 9,911 -6%
29-Aug-08 1,075 9,134 3 10,209 -3%
30-Sep-08 3,174 5,827 2 9,001 -14%
31-Oct-08 1,856 6,149 2 8,006 -24%
28-Nov-08 619 7,976 2 8,595 -18%
31-Dec-08 1,856 7,558 2 9,414 -10%
30-Jan-09 1,856 5,224 2 7,081 -33%
27-Feb-09 2,422 3,500 1 5,922 -44%
31-Mar-09 3,119 3,252 1 6,371 -39%
30-Apr-09 1,374 6,479 2 7,853 -25%
29-May-09 1,983 6,930 2 8,913 -15%
30-Jun-09 2,373 6,928 2 9,302 -11%
31-Jul-09 3,055 7,365 2 10,420 -1%
31-Aug-09 3,733 6,911 2 10,644 1%
30-Sep-09 908 11,352 3 12,260 17%
30-Oct-09 1,967 10,102 3 12,069 15%
30-Nov-09 2,215 10,510 3 12,726 21%
31-Dec-09 2,425 10,217 3 12,642 20%
29-Jan-10 3,323 9,385 3 12,708 21%
26-Feb-10 3,323 9,810 3 13,133 25%
31-Mar-10 166 13,527 4 13,693 30%
30-Apr-10 415 12,834 4 13,249 26%
28-May-10 415 11,205 4 11,620 11%
30-Jun-10 415 9,912 4 10,327 -2%
30-Jul-10 415 10,831 4 11,246 7%
31-Aug-10 415 10,123 4 10,538 0%
30-Sep-10 415 11,604 4 12,019 14%
29-Oct-10 415 12,565 4 12,980 24%
30-Nov-10 415 12,703 4 13,118 25%
31-Dec-10 720 14,249 4 14,969 43%
31-Jan-11 1,444 14,582 4 16,026 53%
28-Feb-11 2,146 13,196 4 15,342 46%
31-Mar-11 2,315 12,674 4 14,990 43%
29-Apr-11 2,315 13,077 4 15,392 47%
31-May-11 2,489 12,496 4 14,986 43%
30-Jun-11 2,489 12,075 4 14,564 39%
29-Jul-11 2,489 12,258 4 14,747 40%
31-Aug-11 3,483 11,766 4 15,249 45%
30-Sep-11 3,483 11,862 4 15,345 46%
31-Oct-11 3,483 13,391 4 16,873 61%
30-Nov-11 3,483 13,515 4 16,998 62%
30-Dec-11 258 16,972 5 17,230 64%
31-Jan-12 775 17,227 5 18,001 71%
29-Feb-12 1,497 16,751 5 18,248 74%
30-Mar-12 2,252 17,132 5 19,384 85%
30-Apr-12 2,529 16,202 5 18,731 78%
31-May-12 2,529 13,867 5 16,396 56%
29-Jun-12 2,529 13,984 5 16,513 57%
31-Jul-12 2,529 13,549 5 16,078 53%
31-Aug-12 1,118 16,180 5 17,298 65%
28-Sep-12 1,118 16,667 5 17,785 69%
31-Oct-12 1,118 16,098 5 17,216 64%
30-Nov-12 1,118 16,698 5 17,816 70%
31-Dec-12 1,118 16,961 5 18,079 72%
31-Jan-13 1,286 17,695 5 18,982 81%
13-Feb-13 1,705 17,754 5 19,459 85%



Overall net profit = 19,459/10,500 = 1.85 i.e 85.3%

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241769 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 9:34 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Correction to 241767:
The Gross profit is 85.3%, the Net profit is 74.7%

Alicia

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241772 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 10:43 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 9
Alicia wrote:
However, there have also been some helpful comments. Incidentally, on the BAC issue my results were correct but in retrospect I have realised that I was using a slightly different 20-day trend function than I had posted.

And also posted results that still contain survivorship bias (crystal ball) picks in her backtest.

Wow! Even I'm a tad surprised. You are now saying that BAC was correct even when I have proven it to be wrong! How? By claiming to have used a different formula, even though the formula shared chose the same picks in the vast majority of the historical dates. Seriously?

I'm with Elan on this... nough said!

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241774 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 10:53 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Overall net profit = 19,459/10,500 = 1.85 i.e 85.3%

You're joking, right?

As the Brit guy said on the TMF-UK board, "You're having us on." This is clearly some sort of joke. But I just don't get the joke -- could somebody explain it to me? Thanks.

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241775 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 12:13 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Zeelotes (241772)

Wow! Even I'm a tad surprised. You are now saying that BAC was correct even when I have proven it to be wrong! How? By claiming to have used a different formula, even though the formula shared chose the same picks in the vast majority of the historical dates. Seriously?

Yup. True

I was using

B40=TREND(A1:A21,,{21})/TREND(A1:A21,,{20})-1
C40=TREND(A21:A41,,{21})/TREND(A21:A41,,{20})-1

AND I'VE DONE IT AGAIN!

For the 30-day trend use:

TrendA=TREND(A1:A30,,{31})/TREND(A1:A30,,{30})-1
TrendB=TREND(A30:A59,,{31})/TREND(A30:A59,,{30})-1

Sorry Guys - the error is in translating my spreadsheet formulae to the MF posting.

Alicia

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241778 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 12:55 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Alicia: Are you even including GM on 3/13/2009?

With your adjusted formula the result is:

BAC = 1.69%
GM = 2.48%

In other words, GM is still chosen, and your system on the Dow still blows up in your face.

This continual changing formulas and parameters to try to tune the historical record to support your hypothesis is really getting tiring.

Print the post Back To Top
Author: MoeBruin Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241779 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 1:01 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
You're joking, right?

As the Brit guy said on the TMF-UK board, "You're having us on." This is clearly some sort of joke. But I just don't get the joke -- could somebody explain it to me? Thanks.


Its kind of like the "Catfish" stuff that was recently in the news. You post a bunch of _____ that you know sounds foolish but will be too tempting to the people reading to ignore and you see how much discussio, in-fighting etc that you can get.

You keep throwing tidbits out there to keep the juice flowing. When the flowing stops you bring the show somewhere else (in this case from the UK to the US board).

Some people get a kick out of getting other people riled up.

Moe

Print the post Back To Top
Author: alevine Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241780 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 1:11 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Overall net profit = 19,459/10,500 = 1.85 i.e 85.3%

You're joking, right?


What am I missing here? What's the joke??


Alan

Print the post Back To Top
Author: heink One star, 50 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241782 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 2:16 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
I "ignore thread"-ed this earlier, but went went back to read it after all after noticing that some of the more well-respected contributors were weighing in (and getting lots of "rec's"). There's been some great discussion, but mainly it's been a reminder that we mustn't get overly invested (emotionally) in our ideas/methods. Remember the de facto MI motto...

I gotta get back to work.

heink

Print the post Back To Top
Author: heink One star, 50 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241783 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 2:22 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Wow, that didn't read at all like I thought it would! The motto was given in: http://boards.fool.com/the-original-poster-asked-a-direct-qu... and I still have to get back to work. Although the idea of watching the market intently while my 2013 gains evaporate is also tempting!

heink

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241784 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 3:17 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Zeelotes (241778)

Alicia: Are you even including GM on 3/13/2009?

Well now I never thought of that.

GM is not in the Dow List I got from Yahoo:

http://finance.yahoo.com/q/cp?s=%5EDJI+Components

That would explain a lot wouldn't it?

Alicia

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241785 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 5:15 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Alicia wrote:
That would explain a lot wouldn't it?

As I thought, you are not reading anyone's responses. You do realize that this sort of thing requires a tremendous amount of work to validate.

Print the post Back To Top
Author: DrBob2 Big funky green star, 20000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241786 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 5:50 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
GM is not in the Dow List I got from Yahoo

GM went into bankruptcy and was delisted from all stock exchanges (and, of course, was dropped from the DJIA). It went public again after a couple of years. The US government still owns some 300 million shares.

DB22

Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241787 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 5:58 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Zeelotes(241785)

As I thought, you are not reading anyone's responses

How come? I read all the responses and reply patiently to those I am able to.

You do realize that this sort of thing requires a tremendous amount of work to validate.

Of course, and it's all in a good cause - to support our understanding, and for the benefit of those less able. I myself have spent untold hundreds of hours designing, developing and testing my code in order to model all the data and to be able to present my case. That is how I am able to respond to challenges so quickly. As I am sure you are aware, it is only by dogged persistence that barriers are overcome and progress moves relentlessly forward.

In terms of furthering understanding and lessening your work load, this link should prove helpful. It will show how the cash at bank ebbs and flows as target buys are processed. It uses the data I presented at post 241767, except that it is better by being continuous by the day rather than monthly.

https://docs.google.com/file/d/0B6rBG1MNaUjJVkZvMTZBVXBid1U/...

And you have not answered my reservation about GM not being in the Yahoo list. It has led to a considerable amount of needless hours by all parties.
Is the Yahoo data correct or not?

Alicia

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241788 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 6:10 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 7
Alicia wrote:
How come? I read all the responses and reply patiently to those I am able to.

And you have not answered my reservation about GM not being in the Yahoo list. It has led to a considerable amount of needless hours by all parties.
Is the Yahoo data correct or not?


In a few posts back I gave you this link:

http://en.wikipedia.org/wiki/Historical_components_of_the_Do...

In it you can easily answer your own question. General Motors is in the list on September 22, 2008 and not in the list on June 8, 2009.

General Motors in fact was in the Dow continuously from 8/31/1925 until it was removed in 2009. In other words, on the date in question in your backtest it most definitely was a component of the Dow.

You ask: Is the Yahoo data correct or not?

Clearly, you do not understand what is entailed to do a backtest like you see on this board. It requires hundreds and hundreds of hours of hard work. Guys like Gritton, Robbie and others have expended huge amounts of their spare time to support this group.

Is the Yahoo data correct? Yes, it is correct TODAY! It may or may not be correct tomorrow because component lists can change all the time. The Dow 30 is one of the absolute easiest lists to backtest since it has only had 55 list changes from 5/12/1924 to the present. In contrast, the Russell 2000 I track changes almost daily.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: AliciaThyme Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241789 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 6:24 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Zeelotes (241788)

General Motors in fact was in the Dow continuously from 8/31/1925 until it was removed in 2009. In other words, on the date in question in your backtest it most definitely was a component of the Dow.

I understand that, but as I have been unable to find any GM data pre 18 Nov 2010, it's not very helpful. And as I already explained in my post 241767 I can only work with available current constituent DOW data (because the only source I have found is Yahoo). Perhaps you can provide a link to historical data of past constituents?

Alicia

Print the post Back To Top
Author: Zeelotes Big red star, 1000 posts Feste Award Nominee! Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241790 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 6:43 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Alicia asked:
Perhaps you can provide a link to historical data of past constituents?

Hit reply to one of my posts and then only check the box:

E-Mail this Reply to the Author

At the bottom. Uncheck Post this Reply to the Boards.

I'll send you a ss with the data you require.

Print the post Back To Top
Author: Alphastotle Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241791 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 7:05 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
Alicia, the problem with doing a backtest with the current list of constituents is that it introduces survivor bias. The stocks that drop from an index are often the ones that have poor performance or gone out of business. In some cases however they have been acquired (probably less likely with the Dow 30) which would be a good thing.

Print the post Back To Top
Author: FlyingCircus Big red star, 1000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241792 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 8:08 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 6
Reminds me of another famous Brit sketch.... the Monty Python cheese shop sketch....

after a cheese shop owner (AT) is asked about 100 times for different kind of cheeses which he does not have,
C: (deliberately) Have you in fact got any cheese here at all?

O: Yes,sir.

C: Really?

(pause)

O: No. Not really, sir.

C: You haven't.

O: Nosir. Not a scrap. I was deliberately wasting your time,sir.

C: Well I'm sorry, but I'm going to have to shoot you.

FC

Print the post Back To Top
Author: elann Big gold star, 5000 posts Top Favorite Fools Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241796 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 9:15 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 5
And as I already explained in my post 241767 I can only work with available current constituent DOW data

Garbage in garbage out.

Why not just define a fantasy stock universe in which all stocks rise 5% every month. You'll make a lot of money. NOT!

Every hour you have spent refining your model and programming it has been an hour wasted, because the first and necessary condition for any test is to have valid data.

Elan

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241799 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 10:30 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
I gotta get back to work.

Heh. The goal of MI is for your "work" to be shuffling around your house in your PJ's until noon[*], taking a break now and then to do some web surfing -- in between playing Angry Birds and Big Kahuna Reef.

Stuff like this thread is just entertainment.

The other bit of work I did this afterniin was watch Argo. Great movie.

------------
[*] Because it's time to go get the mail, and you can't walk up to the mailbox in your PJs.

Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241800 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 10:57 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 1
Alicia: Are you even including GM on 3/13/2009?
Well now I never thought of that.

GM is not in the Dow List I got from Yahoo:
http://finance.yahoo.com/q/cp?s=%5EDJI+Components
That would explain a lot wouldn't it?


LOL!
Yeah, that certainly explains a lot. It explains that you don't have the slightest clue what "point-in-time" means. Or "survivorship bias". Even though people have raised these points and explained it to you several times. I'm not going to (re-)explain them here once again, you can Google them yourself.

Ah heck, why not?
Yup, sure enough, the Dow List Yahoo: http://finance.yahoo.com/q/cp?s=%5EDJI+Components doesn't say that GM is in the DJIA.

But, y'know "backtest" kinda implies that we are testing during other (historical) times. As in "back in time". You seem to think that the DJIA was always what it is today.

Google is your friend. Try something like, oh, I don't know, maybe "djia components history". Which leads to this: http://en.wikipedia.org/wiki/Historical_components_of_the_Do...

Which says that GM was in the DJIA on Sept 22, 2008 but was removed on Jun 8, 2009.
Glancing at a calendar, you can see that 3/13/2009 is *after* 9/22/08 and *before* 6/8/09, so from that you might deduce that GM was a DJ stock on 3/13/09 even though it is not a DJ stock today, 2/21/13.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: Rayvt Big gold star, 5000 posts Top Favorite Fools Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241801 of 253091
Subject: Re: Does this stuff work ? Date: 2/21/2013 11:16 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 2
Of course, and it's all in a good cause - to support our understanding, and for the benefit of those less able. I myself have spent untold hundreds of hours designing, developing and testing my code in order to model all the data ...
All wasted. All that effort a total loss. So sad.
You have been impervious to people's critiques that you have been making serious errors. You have ignored everything that anybody has said about errors or flaws.

And you have not answered my reservation about GM not being in the Yahoo list. It has led to a considerable amount of needless hours by all parties.

Not by all parties. Mostly by you.

I don't know about others, like Zeelotes & Elan, but for me this time has been no more of a waste than an afternoon spent watching movies. That is, it's been pure entertainment.

Okay, nobody else has said this, but it's time for you to re-read all the comments that have been made by others responding to your posts -- both in this thread and on the 2 Brit threads, and then ....

you need to google "Dunning–Kruger effect"

Print the post Back To Top
Author: tetranomad One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241804 of 253091
Subject: Re: Does this stuff work ? Date: 2/22/2013 7:55 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 4
you need to google "Dunning–Kruger effect"

Thanks for that Ray, I hadn't seen it before. Reminds me of my favourite Bertrand Russell quote:

"The whole problem with the world is that fools and fanatics are always so certain of themselves, and wise people so full of doubts."
-Bertrand Russell

Print the post Back To Top
Author: JeanDavid Big gold star, 5000 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241805 of 253091
Subject: Re: Does this stuff work ? Date: 2/22/2013 8:34 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 39
the first and necessary condition for any test is to have valid data.

I spent much of my professional career in computer system design and programming. I even built a relational database management system for UNIX back when UNIX did not have such a thing (late 1970s).

When I got seriously interested in investing, I thought backtesting would be a good way to evaluate mechanical investing methods. I mean, if it did not survive a backtest, I was sure not going to risk real money in a future "test."

So I set up a relational database, containing price, split, dividend, etc., and set up a software structure whereby I could write SQL statements to perform the strategy under test. I thought that would be the hard part, but I was wrong.

The problem was getting accurate data, especially old data. I hoped to go back to about 1970 or so. I had access to some commercial investment data, and Yahoo! price and dividend and split data. I did not have access to merger data, and did not know what to do with it if I could have gotten it.

The Yahoo price data was the worst. Total survivorship failure, for one thing. Prices change every day of course, but tickers change a lot, too. and VL and IBD did not change the tickers at the same date as Yahoo! did. Sometimes several weeks apart. Also, the monkeys at Yahoo keep changing the format of their data so that any mechanical aids to downloading their data and trying to validate it have to be continually monitored and revised.

It might have been beneficial to get the CRSP data, but I was too cheap to do that. After about 5 years I gave it up because with bad data, there was no point, and getting it and keeping it clean took over a year of work for each year's worth of data.

I really admire the work of those who have put a backtester together (the easy part), and getting, cleaning, and validating the data (the hard part) and keeping it all up to date and working.

Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Print the post Back To Top
Author: DoesMIWork One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 241809 of 253091
Subject: Re: Does this stuff work ? Date: 2/22/2013 12:21 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0