These results come with a huge caveat - the time period. I'll have to go through my notes again, but I've done some calculations on the so-called HML premium and found the empirical proof for its existence depends heavily on the "sample window" you use or the markets you examine. The recent evidence is pretty weak, especially for the US markets, but I think it still holds true for some developing markets in the Middle East and Southeast Asia.

I do have a preference for value stocks over glamour stocks because they make sense to me more, intuitively, but I doubt very much that a simple "low P/E (or P/B) screen" works as an effective filter. At best, it's a great place to begin.

Thanks for this article! It brings back fond memories of struggling with datasets..