Are you backtesting with PTF? Try to backtest also with 0.5*stddev. In my opinion, it also makes sense to consider a maxim holding time. Any thoughts on this subject guys?
Pair trading is all about finding cointegration within price series, not correlation. Here's a layman explanation of the difference: http://epchan.blogspot.com/2006/11/cointegration-is-not-same-as.html Amusing enough, we are at page 162 of the thread (very interesing thread whatsoever), and nobody seems to care about this.
No way. It's the third or fourth time, co-integration is mentioned. Personally i trust on backtest results, as a way to measure how strong is the pair. What i have read about co-integration (Johansson, Dickey Fuller ...), is not easy to understand for someone without a statistical background. Also, I doubt it's possible to calculate co-integration in mili-seconds, like you calculate correlation. When you need to analyze hundreds of pairs, the pairs indicators should be calculated on the fly.
careful jonny. not sure how far back you like to backtest, or how far PTF can go, but this spread has not always been so stable. Back in 07/08 the spread ratio tripled (from 1.0 to 3.0) in a couple of months, near straight line then came all the way back even quicker a few months later. only going long OSTK would only have saved you from enduring the 2nd move.
Am not using PTF. I use excel with data fed into it automatically, although would appreciate info on a more all-encompassing piece of software than PTF. With a Gaussian distribution curve the difference in probability from a 1* event to a .5* event is much smaller than that from a 1.5* event to a 1*, and furthermore because these are rolling means, you could wait to trade until it goes back to .5* and still make a loss. The flipside of stop loss/take gain after 1* move is that you often get stopped out early when the stop loss figure is breached, and it also limits your returns. Furthermore, I think there will be more drag from comms. The advantage is that you try to make wins/losses symmetrical, and just try to profit by your overall win rate. This is attractive to me as it makes using money management optimilisation techniques much simpler. It is also psychologically easier because it helps avoid the isolated nasty loss trade. Another thing to consider is that the 1*stdev are not going to be all the same for the trades, however I've found that around taking 2% target (of total invested) works ok. I agree with the maximum holding time argument, as 1/ this is supposed to be a short term trading strategy, based on movements back from short term abnormal pricing 2/ It helps protect you from the fundamental developments that might make the 'abnormal' pricing suddenly appear normal. I'm trying to build up experience of this, and would appreciate observations- that the more experienced guys have-on the above.
You are right, cointegration is difficult to understand and apply, while correlation is a built -in funtion in excel and a lot of other software. But, question remains: why are people assuming that two correlated series are also cointegrated?
I think the trade quality of pairs always changes. Therefore I think a shorter backtesting period (12 months) has more weight than a wider timeframe. For myself I update my list about every 3 months.
i agree. but the trouble with spreads like this one (normally with one stock v small compared to the other) is that they look good until they blow out, then they look terrible. there is no "warning period" where the quality of the spread slowly deteriorates. instead you just get raped as the small stock doubles or halves in value whilst the larger stock does its own thing. either way, wish you luck
Thats the risk you're going to take if you're in the business. Its on you to manage that risk properly. For myself PT is the most stable and less risky approach I found so far and I'm very happy and confident with it.
sure. i'm not a trading (or pairtraiding) naysayer. i trade full-time, including some pairs stuff. risk is all part of the game. my only question is whether pairs like this particular one (ebay / ostk) represent good risk reward. but that's down to the individual i guess.