Home

Mad Em-Dashes is a St. Louis Cardinals blog by Dan Moore that does not want to waste your time. 

Why FiveThirtyEight incorrectly projected Pete Kozma to win the 2012 presidential election.

Last season Pete Kozma did not have one of the most surprising down years since 2006. This is one of the few times I'll feel empowered to disagree with FiveThirtyEight on a math question, so please excuse me for stretching it out to post length.

If you read the link, you'll see that what they're doing has a lot in common with the old PECOTA percentile forecasts, which were always one of my favorite parts of subscribing to Baseball Prospectus. Pete Kozma's .238 wOBA is, according to a chart of disappointments bookended by Travis Hafner and Clint Barmes, at the 0.2 percentile of what we could have expected from him.

Here is the thing: Nobody expected this of him. Neil Paine, the author, is using Marcel, a famously simple projection system that its inventor explained like so:

The Marcel the Monkey Forecasting System (or the Marcels for short) is the most advanced forecasting system ever conceived.  Not.  Actually, it is the most basic forecasting system you can have, that uses as little intelligence as possible. So, that’s the allusion to the monkey. It uses 3 years of MLB data, with the most recent data weighted heavier. It regresses towards the mean. And it has an age factor.
I've always thought of it as a (needed) corrective to the famously Ptolemaic PECOTA; if you feed it three years of data, you get a surprisingly good projection.

But it doesn't have three years of data for Pete Kozma. It has 2012, in which he hit .333/.383/.569 in 82 major league plate appearances. His .548 OPS last year was extremely surprising if your data is a mix of Pete Kozma's 82 best plate appearances ever and "the league mean over 200 PA".

I think/hope that Paine's doing more than that here, but based on his result—Pete Kozma's projected wOBA is .350, about even with Andruw Jones before his own collapse—he can't be.

I don't know where I first read it, or even if I'm paraphrasing it accurately, but I've always been struck by something Bill James said about new statistics: If you're getting a bunch of results that make no intuitive sense, and confirm none of your preexisting hypotheses, you might have a problem. If your pitch-framing research suggests that Tom Glavine was fooling umpires on his deathbed, you might have a new statistic.

Pete Kozma hit .214/.280/.289 in AAA in 2011 and .232/.293/.355 in AAA in 2012.

He hit .217/.275/.273 in the major leagues in 2013. ZiPS projected .226/.281/.328.