Add new comment

The links were indeed interesting.  Though if I were going to sit down and chat with them over crabcakes, my first question would be "What do you see as the pro's and con's of using regression so much, in the context of human baseball performance"?
I've got my reservations, to start with, about assuming that everybody's going to be mediocre next time around (give or take a few points based on whether they're good or bad players).  James' 1977 "Plexiglass Principle" has run amok to the point to where EVERYONE's "correct" projection is based on "regression to the mean."  Would that they still called it the Plexiglass Principle so that we didn't have the misconception that this was hard science.
But the Buck article even selectively "regresses" away his positive recent performance, while leaving his earlier more negative performances intact.  Call it regression and you're licensed to kill, it seems.
With all respect, I don't believe that "we can pretty easily estimate a hitter's platoon skill" but cheerfully acknowledge that the *saber* consensus is that we can.  The LL articles do a good job of applying mainstream saber thinking, in my view.
........
As many sabertistas themselves understand, all the PECOTA/Steamer/etc models do a *terrible* job of predicting breakouts, falloffs, minor leaguers, and anybody who isn't firmly in the middle of the bell curve with Cano-like established levels of performance.  You could predict every single player in the league to simply hit 260/330/420 and get results reasonably similar to what you get from PECOTA.   Dr. D's new blog publishes an OPS "projection" table for 1,400 players:
+20/20/20 - major stars
+10/10/10 - good players
260/330/420 - the 60% in the middle
-10/10/10 - bad players and rookies
-20/20/20 - known terrible players
This "Dr. D 260/330/420 Projection System" would get an R-squared of about 25%, compared to PECOTA's 30%.  What it would be GOOD for, would be a different question ... :- )
 
I strongly suspect that aging platoon hitters would be another example of "odd duck" skill sets that can't just be "projected" towards 260/330/420.
I've got a hunch that the (age-d) Seth Smith has more of a platoon split, rather than less.  Same with Ruggiano.  But who am I to argue with The Mighty Regression Law of Saberphysics :- )
........
Good link Dr. G.