Factset and Alpha Tester

This semester, we were encouraged to get to know Factset's Alpha Tester package. Wow, is this thing addictive (to me). After spending an hour with the online tutorial, I started testing away. I've built so many screens over the years that I have wanted to backtest, I can literally run a backtest per night if I want to.

Thus far, with the tool I have merely figured out many of the nuances such as which database to grab pricing information from etc. Some of my results have also been fairly intriguing. We have long joked that the Price to Sales ratio was pretty useless (Kenneth Fisher swears by it, though). It turns out this may not be the case...because of the factors I tried (23 of them, mind you) from 1987 through 2005 P/S was the most statistically significant explaining excess returns. I think there may be some serious internet bubble bias however...

I also ran a screen written during my internship here that has been shown to (initially) generate pretty strong returns. Alas, buying stocks from the screen results in a really risky portfolio, so the return can be explained, to an extent. I still have more work to do here, however, because I have learned some new Alpha Tester idiosyncrasies.