I agree with those who say cross-project stats make things more interesting and having a reliable set of those would be fantastic. However, we're not there and having project admins change credit on a whim is frustrating. There are always going to be performance gains/losses from project to project because the computations are different. People are going to figure out how to exploit those.

Regardless, what irks me most is when I read, "I'm not going to run Project X because it doesn't score well enough." There are some very worthwhile projects out there which don't attract much CPU time; some score poorly and some are not even Boinc. Stats are great, I love stats, but the goal of any DC project is more than piling up numbers on a webpage. (Seti excluded, of course. )