Who Has Impressed PECOTA The Most?

By Jason Wojciechowski on June 19, 2014 at 9:07 PM

The internet baseball nerdosphere has been inflamed in the last few days by a frankly stupid fight over projections and in-season data and scouting and who even knows what. It's really just been a massive pile of foolishness.

Which is why I want to jump from there to this post!

PECOTA, the Baseball Prospectus projection system, is run in the offseason to produce a seasonal projection. (Really, it produces a range of projections, and then the weighted mean is calculated to get at a sense of the middle of the range.) (None of what I'm saying is going to be particular to PECOTA. I just wanted to pick a system for this post and that's the one I picked.) In-season, a player's projection is updated for the rest of the season ("ROS") by incorporating the new data that we have from in-season performance. PECOTA isn't fully rerun every day because that would be too computationally intensive, take too much time, etc. You can read about how the ROS projections are done here.

A player who does not deviate much from his projection obviously won't see his ROS projection move very much -- if you're hitting .270 and PECOTA said you'd hit .275, well, good job. That's your projection! On the other hand, if you're deviating a lot, the question is how much your projection moves, and this essentially a function of how much data it had on you coming into the season -- if you're a five-year vet, then one third of a season of hitting isn't going to move the needle much because it's about 1/15th of the total data. If you're a rookie, then all your new data is the first major-league data is has for you.

So! Who has moved the needle on their projections the most so far this year? I'll use True Average as a catch-all offensive measure.

Player Preseason TAv Current TAv ROS TAv Difference (ROS - Pre)
Derek Norris .271 .336 .278 +7
Brandon Moss .276 .313 .281 +5
Coco Crisp .266 .310 .270 +4
Nick Punto .234 .250 .236 +2
Yoenis Cespedes .288 .287 .288 0
John Jaso .284 .287 .283 -1
Alberto Callaspo .265 .246 .263 -2
Josh Donaldson .287 .298 .283 -4
Jed Lowrie .271 .246 .267 -4
Craig Gentry .257 .259 .251 -6
Eric Sogard .260 .201 .249 -11

So that's not at all a surprise, right? There's some weirdness in there in the middle, sure, like Josh Donaldson and Craig Gentry dropping points in their ROS projections despite outhitting the preseason line, but ... well, who knows, maybe it's a bug! I don't know what to say. But at the bottom you have Sogard, a player without a long history of performance hitting well below expected, and Norris, a player without a long history of performance hitting well above expected. Those are your two big gainers and losers.