What about Greenblatt's "Magic Formula" and some of the screens discussed in Quantitative Value?

I wouldn't dismiss screening per se; I think it's a matter of using the right metrics. I agree that screening with criteria based on forward estimates (such as P/EG) won't get you too far, but with the right metrics (based on historic rather than forecasted data) I think you can find some good, undervalued companies.

There's also a difference between using a screen as the sole method for picking stocks or using it as a starting point for equities that merit further research. I think the latter approach combines the "art" and "science" aspect of investing, and is certainly better than blindly choosing stocks just because they have a low P/E or P/B ratio, for example.