Bone, An interesting thought. People do pairs trades with one stock against another all the time. But as perhaps you mentioned, you really need a lot of capital to make it worthwhile, and do be able to spread your risk across dozens of pairs. You mention spreading one or more stocks against e-mini contracts. Can you give some examples? Are you just doing stuff like QQQ against NQ, SPY against ES, etc.? Or are you picking other stocks which have high correlation to the index, such as CSCO against NQ, etc.? I'm curious, as I've seen full-blown index arb operations, running full baskets (the entire index) against NDX or SPX futures. I personally don't have the capital for that size of operation, and was curious how it would work out with a small group of them. How do you go about it? Thanks.
I would choose stocks that correlate highly to an index, like GE to the S&P, and Cisco to the NAScrack.
What do you use to find correllated equities? Right now I spend a lot of time backtesting hunches (which actually hasn't been too bad, but I wonder if I'm missing better pairs.) I'd love to find a mechanical scanner.
This is still another area where a Bloomberg comes in mighty handy. It does the analysis for you. As a matter of fact, you want to do this exercise on a frequent basis, as the basket changes. You don't want to do the entire basket, or even a populated basket, against the futures index. The fees and commissions makes it a futile effort. Instead, concentrate on one or a handful of stocks against the futures index. Now, my friend, you're a true arbitrageur. I paid over a million dollars in commissions last year. It's what you give up for consistency.
Bone. Good job in explaining the intricacies of stock v. stock pair trading and it's hidden dangers. Look at HD and Low. The only 2 in the home improvement business so intuitively would be a good "reversion to the mean" spread- has been between $2-7$ for a while w/ HD at premium now it is 6$ under. Anyone hoping for mean reversion would be out of business. Your idea of index arbing is more sensible. I've heard of a somewhat similar idea using options indexes vs. certain component of the index called dispersion wherein straddles of index are arbed against straddles of a couple of its components. Anyone done this. Thanks
Found this old thread.. anyone here doing synthetic pair trading? E.g. some linear combination of stocks moving along with another real (or synthetic basket).