No. of Recommendations: 17
Hello. This post discusses a free cash flow screen with a 12-month hold that since 1997 has a CAGR of 69 and a Sharpe Ratio of 1.38.

I haven't posted here in a while, but have been monitoring the board. Was very pleased to read that Fireballs has picked up the mantle from Brian Finney and Jack Cade.

Six months ago, in Message #112226, I commented on concerns then being raised about the PEG screens. My thought was, while PEG screens should perform well in the long run, that problems can arise because earnings depend on the machinations of corporate accountants. We all know of instances (e.g. PPD, Lernout & Hauspie, Enron) where accountants grossly distorted earnings figures.

I suggested that Free Cash Flow (FCFL) might be a more dependable substitute for earnings because it is a more difficult figure to manipulate, and posted a possible monthly screen.

Thanks to the comments of other posters, I took another look at the possibility that FCFL screens might be more applicable to longer holding periods. This coincided with my own nature which would prefer to invest in a few stocks once or twice a year rather than trading several on a monthly basis.

As a result, last summer I developed three new FCFL screens: one an annual and two semi-annuals. I've delayed writing about them until now as I wanted to to have a much information as possible prior to the January trade date. While the backtester only provides information on these screens since 1997, the fact that they have performed well in both bull and bear markets leads me to feel that they have some validity. This post deals with the annual screen, FCFLAN.

FCFLAN

Note that FCFLAN has no RS component. The URL for this screen is:
http://backtest.org/?SB9701011222XOfcflDcsoDpriCT8XOfcflDcsoDpriCB7XdpcB5UUUU

The comparisons with the S&P are as follow:

                             FCFLAN               S&P

CAGR                           69                 10
GSD                           45                 19
Sharpe                       1.38               0.39

Winning Stocks             19
Avg. Gain                     76
Losing Stocks                 4
Avg. Loss                   - 50

1997                           65                 33
1998                           54                 29
1999                          151                 21
200                            27               - 9
2001                           73             - 13

A few comments on the reasoning behind the development.

1. I worked with TimeinessTM 2-2 because my past investing experience indicated that analysts raise stock ratings to the highest possible level only after superior performance for some period of time. My thought was that potentially successful longer term holdings (at least a year) may be more plentiful in the Timeliness 2 grouping.

2. The FCFL aspect is simply a sort with the 8 highest ranked ratios of FCFL per share divided by stock price, and then eliminating the highest ranking stock. For some reason the highest ranked stock always performed below the average in whatever permutations I tried.

3. The last step selects the 5 stocks with the lowest debt to capital ratios as of the last quarter.

I'll post the information on the two FCFL Semi-Annual Screens when I have some time tomorrow.

As always, any comments or thoughts are welcome.

Slievebricken
Print the post Back To Top
No. of Recommendations: 2
There is an obscene January effect going on. Try some of the other start months.

Larry
Print the post Back To Top
No. of Recommendations: 1
1. I worked with TimeinessTM 2-2 because my past investing experience indicated that analysts raise stock ratings to the highest possible level only after superior performance for some period of time. My thought was that potentially successful longer term holdings (at least a year) may be more plentiful in the Timeliness 2 grouping.

T=2 not only contains stocks recently promoted from T=3, it also contains stocks recently demoted from T=1. You can't rely on it to choose stocks whose best is ahead of them. For many the best is in the rear-view mirror.

- Joe
Print the post Back To Top
No. of Recommendations: 4
While the backtester only provides information on these screens since 1997

1. I worked with TimeinessTM 2-2 because my past investing experience indicated that analysts raise stock ratings to the highest possible level only after superior performance for some period of time. My thought was that potentially successful longer term holdings (at least a year) may be more plentiful in the Timeliness 2 grouping.

2. The FCFL aspect is simply a sort with the 8 highest ranked ratios of FCFL per share divided by stock price, and then eliminating the highest ranking stock. For some reason the highest ranked stock always performed below the average in whatever permutations I tried.


This looks like a canary killer. The short backtest makes it easy to find spurious results. The Timeliness 2 preference is suspicious. And the elimination of the top FCFL stock is data mining almost beyond a shadow of a doubt.

The comment from winker clinches it -
There is an obscene January effect going on. Try some of the other start months.

I wouldn't touch it with a ten foot pole.

Elan
Print the post Back To Top
No. of Recommendations: 1
What's with all these screens that are backtested on the biggest stock market bubble in the history of civilization? You're going to get some mighty wacky reults on a screen that's only tested back to '97.
Print the post Back To Top
No. of Recommendations: 1
<<What's with all these screens that are backtested on the biggest stock market bubble in the history of civilization? You're going to get some mighty wacky reults on a screen that's only tested back to '97. >>

Since 1997, the stock market has averaged 10% annually.

Since the mid-80's, the stock market has averaged 14% annually.

I disagree that a backtest since 1997 represents a "bubble" period (as a whole). If it were a bubble period, the returns would have been materially higher than the markets historical average.

rick
Print the post Back To Top
No. of Recommendations: 3

E-mail message
Thank you to all those who commented and criticized the new FCFLAN screen. Here are some thoughts on your responses.

1. From Winker:

  "There is an obscene January effect going on. Try some of the other start months."

Absolutely. FCFLAN worked obscenely well only with a January start date. One wouldn't use it for any other month's start date. Isn't that fine so long as its performance in both good and bad years continues? It's strictly a one time per year screen.

On a side note, one of the things I like about the screen is that its best performance seems to take place not in January but during the last months of the calendar year hold.

2. From Mainiac Joe regarding my comment on why I worked only with TimelinessTM 2-2:

"T=2 not only contains stocks recently promoted from T=3, it also contains stocks recently demoted from T=1. You can't rely on it to choose stocks whose best is ahead of them. For many the best is in the rear-view mirror."

Agreed. However, don't stocks that have been demoted also recover? Some technicians use the "cup and handle", others call it retracement followed by consolidation prior to beginning a "new leg up." John Bollinger has written that in his rating system the best moves are often made by those that retrace from a Potential 5 to a Potential 4. Likewise, I seem to recall Jon Markman making similar comments about MSN's Stock Scouter system.
In any event, the purpose of the FCFL sort is to identify from this pool the best values at that point in time. The debt to capital sort screens for financial strength and removes some risk.

But, you are definitely right: Timeliness2, by itself, cannot be relied upon to choose stocks whose best is ahead of them.

3. From Elann:

"This looks like a canary killer. The short backtest makes it easy to find spurious results. The Timeliness 2 preference is suspicious. And the elimination of the top FCFL stock is data mining almost beyond a shadow of a doubt.
The comment from winker clinches it -
There is an obscene January effect going on. Try some of the other start months.
I wouldn't touch it with a ten foot pole."

I cannot disagree with your comment about the short backtest. I wish it was longer, and only presented it because of its performance during both bull and bear markets.

The elimination of the top FCFL stock is not necessary for it to still show good results from 1997-2001. If one alters the screen to Timeliness2; selects the 10 with top FCFL per share divided by price; and then selects the 5 with the lowest debt to capital ratio, it still shows a CAGR of 53, GSD of 42 and a Sharpe Ratio of 1.20.

http://backtest.org/?SB9701011222XOfcflDcsoDpriCT8XOfcflDcsoDpriCB7XdpcB5UUUU

Once again, thank you all for your thoughts, and let me apologize for any typos or other problems. My WebTV Classic has been particularly obstreperous these past few days, and it has been a horror to cut and paste or edit.

I will post the two semi-annual FCFL screens either later today or early tomorrow.

Slievebricken
Print the post Back To Top
No. of Recommendations: 0
I cannot disagree with your (= elann's) comment about the short backtest. I wish it was longer, and only presented it because of its performance during both bull and bear markets.

This sounds reasonable, but, alas, isn't. You don't have a short backtest, you have an ultra short backtest, because of your 12 month hold. That gives 5 (five) data points. That is not a backtest, but sheer randomness. What happened in the real world doesn't matter, and can't validate the 'backtest'. To test for robustness, I changed this to a 6 month hold and got 19/36/0.55 (instead of 69/45/1.38 for the 12-month hold). While it is still better than the market, the difference stinks to heaven.

Let me put it like this: it's your money. To all others: don't try this at home.

Regards,

Karel
Print the post Back To Top
No. of Recommendations: 4
Absolutely. FCFLAN worked obscenely well only with a January start date. One wouldn't use it for any other month's start date. Isn't that fine so long as its performance in both good and bad years continues?

I think most people on tghe board would agree when I say no. The best indication of the true historic performance of a screen comes from averaging it s returns over all starting months (or all starting weeks if you have the data). The greater the variation among starting times the less reliable (more data mined?) the screen.

To explain, let me go back to red and white marbles and random sampling. Let's say you picked five marbles out of a bag and got all white. You repeated the exercise 11 more times and got various results. The average of all results was 3 white marbles and two red. Here are two possible conclusions:
1. The ratio of white to red marbles in the bag is probably close to 3:2.
2. You have a good chance of picking five white marbles on your first try, but not on other tries.

Which conclusion sounds more plausible?

Elan
Print the post Back To Top
No. of Recommendations: 2
To explain, let me go back to red and white marbles and random sampling.

Uh oh, it's those marbles again. You better watch out - Elan's never wrong when it gets down to the marble level.

Here are two possible conclusions:
1. The ratio of white to red marbles in the bag is probably close to 3:2.
2. You have a good chance of picking five white marbles on your first try, but not on other tries.
Which conclusion sounds more plausible?


Exactly. Trust Elan when marbles are at issue :-). You've likely found nothing more than a mirage, and we're trying to do a favor by poking holes in it.

- Jamie
Print the post Back To Top
No. of Recommendations: 1
Print the post Back To Top
No. of Recommendations: 0
The Canarykiller is not bad... indead !
http://backtest.org/?SB9701011222XOfcflDcsoDpriCT8XOfcflDcsoDpriCB7XdpcB5UUUU


No reason to throw out the #1 Price/FCF stock, the CAGR only declines a point or two when you do. Maybe make it top 10 by Price/FCF then final sort by the second criteria, to address datamining concerns: then show results for all start months. I like that it's a value screen for annual hold. It's different from some of the other canaricides posted.
Print the post Back To Top