Ok, so you propose that you should have a number for how many times that the game has been played. And maybe only allow ratings for games that has been played more than a certain number of times.

This sounds a lot like the BookMatcher (TM) and Moodmatcher(TM) stuff that Amazon.com has or the CDnow or Music Boulevard recommendations center. I would guess that the design of rudimentary engine would take a weekend and that the implementation would take a few weeks. Clearly there are many different ranking schemes available (preferential, ranked preferential, whatever else they taught me back in poly sci 101...), moreover there are many different scoring schemes that could be implemented with ever-increasing complexity (difference measure, distance measure, cluster analysis, and normalized stuff, subclass analysis based on attributes, various neural network training stuff, et al.). From prior experience, I would say that the mechanism used for making recommendations should be blind to the user.

What's important is whether the recommendations are working. So if someone implements the project, I'd say implement a simple system first and have a feedback page. If the recommendations aren't very satisfying, implement a more complex recommendation scheme. This is an interesting project that would combine both database design/ implementation and some medium complexity cgi. If anyone is interested in working on it, I'd be willing to participate in db design and some of the scoring algorithm development.

One comment on scoring criteria. There are several orthogonal criteria by which a game can be judged. These include complexity, time investment, replay value, appearance, 'fun factor', etc. I mention this just to say that I don't agree that number of times played is always an important criterion. (My specific example is that most of my gaming group does not enjoy economic games while I like them quite a bit. I have my preferences about which games I enjoy the most but I cannot get a group of four players to play 1830 very often, so I've played it something like three times.) For this purpose, I think that multiple scoring schemes should be used (or at least discussed).