Originally Posted by TRF
If I understand PECOTA correctly (and it's very possible that I don't), there are zero human adjustments made to the projections. It's all done by computer.Originally Posted by Falls City Beer
Say the player is a 25 year old Adam Dunn. The computer finds players that were most similar to Adam Dunn up to this point in his career (for example, 25 year old players, similar height/weight, similar OPS, etc). The program then analyzes what happened to that class of players the following year. For example, 60% may have declined, 25% may have collapsed, 20% may have improved.
It's simply a tool. Not perfect (for example, you have to look at the list of similar players and see if there are any mitigating factors... like perhaps one of the "similars" broke his leg the next year), but pretty damn impressive, if you ask me.
While Pecota did miss on Pena in 2005, it also projected his breakout season in 2004. It projected Johnny Gomes' last year.
Pena's four most similar players going into last year were Albert Belle 1990, Jesse Barfield 1983, Bob Watson 1969, and Dale Murphy 1979. 14 of his 20 most similar players trended upward. That's why the program projected improvement in '05. It just didn't happen.