This is my first post in the community. I did a little bit research on Mean Reversion and found the papers about OLMAR and RMR algorithms (this and this). The idea behind those algorithms are simple, it forecasts the future price by taking the mean (OLMAR) and L1-median (RMR) of the windowed historical price. So I extended the idea to forecast the future price by taking x-quantile of the windowed historical price (I use x=0.8).
I tested it to trade on stock list of Amazon, Apple, and Microsoft from 1/1/2011 to date. The result is pretty good I think. I benchmarked it as well with Amazon only (the highest return among the 3) and it still gives more return than the Amazon stock.
The problem is that the algorithm is quite sensitive on what stock I chose. Adding GOOG to the list makes the result worse. I guess some other algorithms also have this kind of problem. Does anyone have an idea on a rule-of-thumb way to choose the stock list for mean reversion technique?