UnThreaded | Threaded | Whole Thread (3) | Ignore Thread Prev Thread | Next Thread
Author: tedthedog One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: of 252783  
Subject: Things that go bump in the night Date: 12/12/2012 4:53 PM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 13
I've posted a sort of fun analysis of day trading versus night trading for gold and for SPY at this link:
http://dl.dropbox.com/u/52242280/DayTradeNightTrade/DayTrade...
Click on the above link for the version with graphics.



#-----------------------------------------------------------------------
Only come here if the above link is broken.
If the link above is working click on it and ignore the following.
Below is the text of the link, including statistics, but w/o any graphics:


INTRODUCTION:
Consider two (very active) trading strategies;
(1) an intraday strategy: buy-open/sell-close
(2) an overnight strategy: buy-close/sell-open

Questions:
(a) Is night-trading better than day-trading?
(b) Are more gains made at night (overnight) than in the day (intraday)?

Disclaimers:
I saw some references to the intraday versus overnight effect on the web and decided to investigate.
Please note that I am not suggesting that anyone try to trade any strategy discussed below.

Googling "intraday versus overnight", or similar, brings up a number of links referencing this effect in gold, equities, and currency trading.
Here's one that discusses SPY: http://marketsci.wordpress.com/2010/09/30/overnight-vs-dayti...
Here's one for GDX: <A http://www.optionetics.com/marketdata/article.aspx?aid=24606...
Here's one from some gold&silver options newsletter company who seem to take it seriously (more on that later): http://www.zerohedge.com/news/overnight-longintraday-short-g...


ANALYSIS:
I programmed up some analysis in R, borrowing liberally from
<A href="http://ennlightenment.com/2012/12/looking-at-gold-stock-intr...

Here are some plots describing the intraday versus overnight strategy, and buy&hold.

First plot: SPY cumulative returns for intraday, overnight, and buy&hold (percent of initial investment at given date)
Second plot: SPY overnight strategy 12 month rolling CAGR (compound annual growth rate over past 12 months, percentage)
Third plot: SPY overnight strategy drawdowns (drawdowns, percentage)
No friction.

<IMG src=SPY.overnightplots.gif alt=SPY.overnightplots.gif>
Here are some associated statistics:
CAGR.overnight Annualized Return 8.68% MaxDrawdown.overnight -34.76%
CAGR.intraday Annualized Return -2.37%
CAGR.buy&hold Annualized Return 6.1%

CONCLUSION:
For SPY there's a moderate edge to night-trading over day-trading and over buy&hold.
More interesting is this general observation:
Bull markets tend to run in the night, not in the day, i.e. the gains of SPY tend to be made overnight (green curve, top plot).


Here's a similar analysis for gold, GDX (no friction):
First plot: GDX cumulative returns for intraday, overnight, and buy&hold (percent of initial investment at given date)
Second plot: GDX overnight strategy 12 month rolling CAGR (compound annual growth rate over past 12 months, percentage)
Third plot: GDX overnight strategy drawdowns (drawdowns, percentage)
No friction.

<IMG src=GDX.overnightplots.gif alt=GDX.overnightplots.gif>
Here are some associated statistics:
CAGR.overnight Annualized Return 56.45% MaxDrawdown.overnight -21.09%
CAGR.intraday Annualized Return -34.11%
CAGR.buy&hold Annualized Return 3.07%

CONCLUSION:
For GDX (gold) there's a huge edge to night-trading over day-trading and over buy&hold.
Again, bull markets tend to run in the night, not in the day.



If the effect is real, why isn't it traded away?
One answer may be "friction".
A little googling suggests that 0.1% might be a reasonable guess at bid/ask spread for quite liquid ETF's (may be on the low side).
I applied the spread to the open and the close data, i.e. each daily/nightly return was reduced by a factor of (1-0.001)*(1-0.001)=0.998
For the daily/nightly strategies this gets compounded daily/nightly.
For buy&hold the 0.998 was applied two times, once at the open and once at the close.
If someone has a better number, or better way to model it, I'd be happy to use it.
But results will presumably be pretty sensitive to the spread if the percentage lost to bid/ask spread is somewhere near the daily percentage gained.

Back to SPY with friction.

Here are some plots describing the intraday versus overnight strategy, and buy&hold.
First plot: SPY cumulative returns for intraday, overnight, and buy&hold (percent of initial investment at given date)
Second plot: SPY overnight strategy 12 month rolling CAGR (compound annual growth rate over past 12 months, percentage)
Third plot: SPY overnight strategy drawdowns (drawdowns, percentage)
With 0.1% bid/ask spread.
<IMG src=spy.withfriction.gif alt=spy.withfriction.gif>
Here are some associated statistics:
CAGR.overnight Annualized Return -34.35% MaxDrawdown.overnight -99.97%
CAGR.intraday Annualized Return -41.02%
CAGR.buy&hold Annualized Return 6.1%

CONCLUSION:
Bid/ask spread of 0.1% kills any gains of the night-trading strategy for SPY.


Back to GDX with friction.

Here are some plots describing the intraday versus overnight strategy, and buy&hold.
First plot: GDX cumulative returns for intraday, overnight, and buy&hold (percent of initial investment at given date)
Second plot: GDX overnight strategy 12 month rolling CAGR (compound annual growth rate over past 12 months, percentage)
Third plot: GSZ overnight strategy drawdowns (drawdowns, percentage)
With 0.1% bid/ask spread.
<IMG src=gdx.withfriction.gif alt=gdx.withfriction.gif>
Here are some associated statistics:
CAGR.overnight Annualized Return -5.51% MaxDrawdown.overnight -52.61%
CAGR.intraday Annualized Return -60.21%
CAGR.buy&hold Annualized Return 3.07%

CONCLUSION:
Bid/ask spread of 0.1% kills any gains of the night-trading strategy for GDX (although there are positive rolling CAGRS during the recent financial meltdown).


Parting shot:
Considering the fricitionless results above again, the top graph in the group of three suggests an obvious short strategy:
Short the day, long the night.
This really amps up the returns if there's no friction.
But it dies when implement a 0.1% bid/ask spread.
Could there be some variation of the above that might work?
Possibly, but it seems pretty sensitive to assumptions of the back-test.
One would have to be pretty confident (or foolish?) to trade any significant funds based on back-tested results of variations of this strategy.
Perhaps the most interesting thing is simply the observation that gains tend to be made in the night, not in the day.
<HR>


I borrowed a lot of the R code to produce the above analysis from http://ennlightenment.com/2012/12/looking-at-gold-stock-intr... (R is open source).
My hack of it is "DayNightTrade.R".
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
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: 240577 of 252783
Subject: Re: Things that go bump in the night Date: 12/16/2012 7:43 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
From a trading cost perspective this might be more feasible than you imagine.
There is no bid/ask spread on MOO and MOC orders, only commissions.

The problem that I see is that to some unknown extent this is a test
of the accuracy and/or biases of reported opening and closing prices.
Do they match the execution price you would get with MOO and MOC orders?
Maybe that's a tiny issue, maybe not.

Of course the other bigger issue is whether this pattern will continue in future.
I had a quick look at GLD and the strategy seems to have worked perfectly until summer 2011.
But a "long at night" portfolio would be down a bit since last August,
matching the net returns of GLD itself almost perfectly.

Jim

Print the post Back To Top
Author: tedthedog One star, 50 posts Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 240582 of 252783
Subject: Re: Things that go bump in the night Date: 12/16/2012 9:30 AM
Post New | Post Reply | Reply Later | Create Poll . Report this Post | Recommend it!
Recommendations: 0
There is no bid/ask spread on MOO and MOC orders, only commissions.
Thanks, I wasn't aware of that.

Do they match the execution price you would get with MOO and MOC orders?
Good question. Don't know if any such stats are readily available, it seems pretty specialized info.
Even so, it's not a strategy I'd personally pursue because it's so active.

I had a quick look at GLD and the strategy seems to have worked perfectly until summer 2011.
Agreed. How well it works in backtests depends on the ETF (or equity, it's fun to see what happens with e.g. MSFT etc).

I did the analysis mainly for fun. It makes one wonder though,
how many other odd market behaviors exist (and why),
and if they're tradeable (if only by the big firms)?

Print the post Back To Top
UnThreaded | Threaded | Whole Thread (3) | Ignore Thread Prev Thread | Next Thread
Advertisement