The best site to use is...any 1 of them. Chone did an analysis and found the correlation coeffecient R below:
.445 Baseball Info Solutions
http://lanaheimangelfan.blogspot.com/20 ... stuff.html
Tangotiger determined that the max correlation was .73
http://www.insidethebook.com/ee/index.p ... /#comments
(However, Pecota did beat it with a .736 but Chone chalks that up to random variation)
Keep in mind that "marcel" is just a simple regression to the mean...not really a forecast per se, just a 'dumb monkey's forecast' and it did pretty well against "the experts."
You learn a few things:
1) nobody is perfect
2) they ain't all that different from each other
3) it's easier to project hitters than pitchers (ergo, pitchers are more risky)
4) they are all pretty good at hitters and all pretty bad at pitchers
Whoever was in 1st place last year is likely to be bumped by somebody else this year. You'd need a lot of years of studies to figure out who "the best" is.
As Ender said, it depends on how much you work you want to do. Pick a system, any one, and then pay attention for the next 6 weeks and do research--tweak where needed.