Pair Trading Strategy Journal

Quote from amitkumar_ny:

I noticed v303 calculating differently too. I asked Jared and he said they modified the code to only look at 1 year data in v303 to avoid software from crashing. He thinks this number is better or more informative. I don't know what to make of the whole thing. I am running DFuller tests in matlab as well and it does seem that cointegration moves a lot based on near term data. e.g. SAIA / XTN was a great pair and both CADF and PTF numbers calculated higher COINT. Recently (past couple of weeks) the pair has gone to hell and CADF test says they are not co-integrated any more. PTF v301 agrees and shows 13% cointegration. So it works but if you get in on a pair and it goes to fell and cointegration number falls off do you get out ? or wait to see if comes back. Half life of this pair is like 12 days so do you wait ?



That is the same question that I have. Wait or not wait - after the co-integration value of a pair plunge. From my prospective if the value fluctuates wildly from day to day then the backtest for finding pairs becomes irrelevant and unreliable. The PTF pairs determined today will not be the same tomorrow if the of value of the pairs fluctuates dramatically. After all we are looking for pairs that move in almost tandem fashion for a duration of time to capture the discrepancy when they are of whack. It also means you need to re-run the back test every week or so to determine the new pairs. The value will change but should not be dramatic. From my understanding for the value to have meaning the look back day should not be too long and not too short. I don't know what the optimal value should be. Maybe 90 days to 120 days?

After all, there is a discrepancy in v301 and v303. Now I wonder why I was told they should be the same. BTW, how responsive is Jared with your questions? I had to wait for almost a week to get a response (that is after repeated emails!). Hopefully, this time I do not have to wait a week again. Needless to say it is frustrating.
More...
 
Question to those using PairTrade Finder.

Which version are you using now? PTF v301 or v303.

Also, did you re-create the database after upgrading?

Thanks.
BT
 
I wish I could understand how COINT is calculated but my brain is not that sharp. I have tried to read some books on it and the amount of matrices they describe makes my head spin. So in absence of true understanding I have to treat this as a black box and take my chances. The more I use PTF the more flaws. I think it should not be used for CI at this point. Use for monitoring price ratios and when it gives a signal use DFuller test or other test to confirm before making the trade....

By the way, do you guys know of any batch yahoo eod downloader. I want to run CADF on yahoo data to match up CI from PTF. Also is there a way to batch convert csv to xls you know of ?
 
I also upgraded to PTF v303. In PTF v301 almost all of my pairs where having cointegration > 0.90. After upgrading to PTF v303, half of my pairs are < 0.90, the lowest pair being 0.04. I wish there would be some release notes to explain what has changed in PTF, and some explanation of why the changes were done.

Has anyone started comparing backtest results between PTF v.301 and v.303? Are the results actually better? Guess I have to make some testing myself, but it would be nice to hear somebody else's experience on this matter.
 
As Amit mentioned earlier the CI calculation is shortened to 100 days (as confirmed by Jared) in v303. I already did an extensive backtest using v303 but I did not have the oppty to compare it with v301. Also, my backtesting criteria is not the same in both versions thus it makes it difficult to compare. In the meantime, I am determined to move on to v303 within the next few days when I close most of my existing positions that were opened using v301. Most of these existing pairs didn't make the cut in v303.
 
Fellow PTF users: Anyone seeing something funny this morning?

This morning I am getting totally skewed quote from Yahoo in PTFv301. Basically, the price for all the quotes are in the $5million range. Needless to say it is giving an entry and sell signal for all the pairs. I think either Yahoo modified its free quote API over the weekend or something wrong with its free quote feed. And if it is an API change, hopefully, a PTF fix is on the way!

Maybe it is time to use IQ Feed's real time quote? Frankly, I don't think there is a big difference with the free Yahoo quote since I always wait a little longer before entering the trade.

-BT
 
Quote from BornToFail:

Fellow PTF users: Anyone seeing something funny this morning?

This morning I am getting totally skewed quote from Yahoo in PTFv301. Basically, the price for all the quotes are in the $5million range. Needless to say it is giving an entry and sell signal for all the pairs. I think either Yahoo modified its free quote API over the weekend or something wrong with its free quote feed. And if it is an API change, hopefully, a PTF fix is on the way!

Maybe it is time to use IQ Feed's real time quote? Frankly, I don't think there is a big difference with the free Yahoo quote since I always wait a little longer before entering the trade.

-BT
More...

I'm having the same problem. Sent an email to tech support already
 
Yes, same problem here. Get a lot of entry/exit signals with some quotes being ridiculously high. I thought it was a bug in PTF v303 as it happened at the same time I upgraded, but maybe that was a hastened conclusion.
 
Good to know it is not just me. I tried PTFv303 too and it is the same. Highly likely that either Yahoo modified the API over the weekend or something is wrong with its quote feed.

I guess I need to call it an early day with pair trading today.

Thanks.
 
Since I have some time on I hand I would like to ask a question for those of you who had been pair trading for sometime for some advice or at least share your thoughts if you are so gracious.

Do you enter trades right away, EOD, or at market opens, once a signal is given during the day?
Or do you wait to see how the wind blows first?

I am asking this question because I notice 'most' trades hit the loss column (unrealize) for the next 1-2 days after a signal is given and then it bounces for a few days like a coil between gain and loss. If this is the case why not enter it after it becomes a loss (let's say after it reach a 2% loss first). I understand some trades do become profitable right away but that is far from the norm.

After all, is patience the main key to pair trading after identifying your pairs?
 
You must log in or register to reply here.
Back
Top