Pair Trading Strategy Journal

[cipherscribe]

Quote from waltbx:
Cointegration seems to be more significant.

Any thoughts?
Walt B

Some good information to think about. I guess it really comes down to whatever works for the individual. Certainly cointegration seems to be the valid method from a mathematical perspective, but correlation seems to be a reasonable facsimile, going by the results of Jonny's trades.

What methods do you use to identify cointegrated pairs?

 

[total_keops]

New trade
Long RY:TD @ 0.7931

 

[fishing]

Quote from waltbx:

Correlation has not been a valid filter criteria in my trades.

I've been pair trading with real money since March 5 using PTF. I've completed 66 trades averaging 3% ROI per week, except for the last week which has been a bummer.

The average correlation of the pairs in my 44 profitable trades has been 57%. And for my 22 losing trades has also been 57%. I do not consider correlation in my choices, either when setting up my pairs, or when entering the trades. Five of my profitable trades had negative correlations at the time I placed the trade (for returns of 18%, 8%, 6%, 6%, 3%) . All of my losing trades had positive correlations. Correlation does not seem to be a factor in profitability.

My initial pair choices are made using profit history. I look for pairs that have a history of over 90% profitable trades in the last year. Correlation is irrelevant, as I see it. I do choose pairs that are in the same industry.

Cointegration seems to be more significant. Problem is, it is far easier to find pairs that are correlated, than cointegrated. But correlation does not seem to be a factor in the 66 pair trades I've made since March 5.

I know this is radically different from the basic premise of pair trading. But after 66 pair trades, I think my statistics have some validity.

Or perhaps there is something unique about the limited period I've been trading?

Any thoughts?

Walt B

thanks for sharing your interesting ideas

could you elaborate more on your entries and exits?

 

[waltbx]

Quote from cipherscribe:



What methods do you use to identify cointegrated pairs?

I don't. I don't know how to identify cointegrated pairs, other than to see a consistent profit history and nice consistent ratio charts. That's as close to cointegrated pair finding as I get. I look for pairs with a profitable history in an industry and choose those to populate my groups.

Daily I scan for pairs diverging more than 1.5 standard diviations from the mean. I don't wait for "signals" and have found many profitable trades that did not trigger a 2 SD signal.

I analyze the ratio chart. When I see an interesting chart, I plug the values into a spreadsheet and check for potential profit. If the potential is greater than 2.5%, the pair is qualified. Some great looking charts (some recommended here) have a potential of only a percent or less. I don't want to tie up my money nor take risk for a low potential profit. I check for news: earnings reports and dividend x dates, other news.

If all seems good, that is a pair I may trade.

Walt B

 

[cooper1308]

Well i've just spent the last 3 days reading this thread from post one

nuggets of gold EVERYWHERE. Thanks Johnny and the rest for everything

I have a question for Johnny,

How much emphasis can I put on winning %? Often i'll find stocks that correlate 40-50-60% however have a win ratio of 97-100% on over 12 trades! with win/loss ratio of 2+

Obviously the more supporting indicators the bigger the edge but how legit is this winning % coupled with win/loss ratio?

If I were to build a universe with say 500 pairs with winning % of 95% and win/loss 2+....would I be correct in allowing for pairs with lower correlation?

 

[Dr Who]

Waltbx, whilst I tend to agree with you regarding your feelings towards correlation (and in fact Rob Friesen agrees with you in his manual), I really don't think you should be getting too excited about such a small sample. 66 trades is hardly a lot and its easy to be fooled by data from this few. Wait till you have a few thousand results.

 

[Ms Varima-Garch]

Jonny, reading your journal with great interest! i'm totally new to spreads

question: when you find a good opportunity, for example, a 2 standard deviation reversion from the mean, and it's a good,
non-trending spread.

how does this play out vs institutional players? . . . i mean does an opportunity disappear quickly because there are a lot of arbitrageur hunters out there?

how quick to you have to be in taking a trade, once you've found an obvious opportunity??

i hope the answer is: relax, everyone can profit

thanks!

Varima

 

[gkishot]

Quote from Dr Who:

Waltbx, whilst I tend to agree with you regarding your feelings towards correlation (and in fact Rob Friesen agrees with you in his manual), I really don't think you should be getting too excited about such a small sample. 66 trades is hardly a lot and its easy to be fooled by data from this few. Wait till you have a few thousand results.

I agree with your observation. 66 trades is a very small sample. It would be much more helpful if PTF provided the profit statistics based on their trades history.

 

[total_keops]

I found an excel spreadsheet that seems to look for cointegration.
I did not test it as I have no plan on using it.
I also found a Matlab code that I will try to use. Again, I did not test the m file yet.
And I am ont sure the Johansen procedure is the right one for pairs trading!
For those interested:
http://www.spatial-econometrics.com/coint/johansen.m

 

[Appleseed]

Quote from cooper1308:

Well i've just spent the last 3 days reading this thread from post one

nuggets of gold EVERYWHERE. Thanks Johnny and the rest for everything

I have a question for Johnny,

How much emphasis can I put on winning %? Often i'll find stocks that correlate 40-50-60% however have a win ratio of 97-100% on over 12 trades! with win/loss ratio of 2+

Obviously the more supporting indicators the bigger the edge but how legit is this winning % coupled with win/loss ratio?

If I were to build a universe with say 500 pairs with winning % of 95% and win/loss 2+....would I be correct in allowing for pairs with lower correlation?

I may be entiely wrong in this asumption but i'll go ahead.

I have a feeling that the winning % is just curve fitting. You can
find hundreds of pairs that are in the winning 80 to 90 percentile
today and going foreward will not maitain that level of a winning percentage of the past. If you select a group of winning pairs today and go forward 6 months with the same consistancy of 80 -90% winners you may have good candidates for pair trading.
Any thoughts??

cheers
john