wilmott.com

Hi Dawg, I got some questions after reading some pairs trading papers recently.Some paper says to use Engle-Granger two step to regress y on x to a get a beta, and then trade with y-beta * xI have two questions about this:1. This beta generally will not give you market neutral, which means that you will have some systematic risk, what to deal with that? A side question is that one want to be somewhat neutral, either market neutral or dollar neutral or someother stuff neutral? I can understand market neutral, but what's the point of dollar neutral? Spliting the funding equally on long and short to maximize funding usage?2. Engle-Granger says test stationary after regression. Again what's the point? I think first of all, the residual most likely is not stationary giving you are regressing price on price, and the beta one gets is questionable. But I am not saying you cannot trade profitly from the residual (go with the local mean?), I just think it will revert back and forth in months. we can just plot the residual instead of using those formal testing, don't we?So, given an example say XOM and CVX, how do we get a beta and form a residual?1. Engle-Grangle say regress XOM on CVX or CVX on XOM2. market neutral says use ratio of the two CAPM beta of the two stocks?Which one should we use? I am confused. A lot of random thoughts.

Dollar value neutral is essential, else one of your legs will be biased. For instance, a tick movement on one of the contracts is about $5 where as on the other exchange it is $12.5, then an approximate dollar neutral would be 5 lots of the contract with $5 tick and 4 lots of the contract with $12.5 tick. (assuming about 0.9 correlation). I am talking about E-mini dow and Emini S&P 500. Check out the CME site: http://www.cmegroup.com/trading/equity- ... _Final.pdf