February 19, 2013
PHOENIX — On Monday, Dave Cameron of Fangraphs posted about the Chicago White Sox consistently beating statistical projections, specifically those of the popular PECOTA model at Baseball Prospectus, which was introduced by Nate Silver and refined by Colin Wyers. Inspired by Bill Pecota, an average ballplayer for the Royals in the 1990s, PECOTA stands for Player Empirical Comparison and Optimization Test Algorithm.
Since 2005, the White Sox have outperformed PECOTA by 57 victories. In 2012, the White Sox won 85 games when PECOTA predicted they'd win 77. In other seasons since '05, PECOTA has been off by 19, 12, nine, seven and six wins on the White Sox. It's gotten close twice, in 2007 (-1) and 2011 (-3).
Cameron concludes that PECOTA doesn't account for the extraordinary health of White Sox pitchers. Over the past 10 seasons, they have lost just under 2,000 days to the DL. Most teams are over 3,000.
PECOTA predicts 77 victories for the White Sox in 2013. I asked White Sox GM Rick Hahn at Cactus League media day how aware he was of these projections, if PECOTA was selling his team short again, and if he knew why.
"It’s not a magic formula by any stretch. If they come up with a 77-win performance, you’re not supposed to take it to the bank," Hahn said.
Wyers notes that "PECOTA has margin of error of roughly eight wins." Even a "perfect" projection model might have a seven-win difference. Some of that is accountable by understanding that no system is perfect. But there's something about PECOTA that gets the White Sox (and the Angels, but let's not talk about them now) wrong.
Hahn isn't certain if PECOTA has trouble accounting for the White Sox only because of their healthier-than-expected pitchers — but it's probably where it starts.
"We tend to outperform that specific system on an annual basis because we do disagree on their projections on our starters in terms of durability," Hahn said. "And this goes back to [Mark] Buehrle, who was one of the more dependable guys year in and year out. But for whatever reason, their algorithms had him at lower inning results than he was yielding."
Simply acknowledging the existence of PECOTA, and that Hahn reads the work of Wyers and others at places such as Baseball Prospectus and Fangraphs, indicates how much respect they deserve. Hahn wouldn't waste his time, otherwise. But some of their opinions do irritate him a little.
"This year, for example, they have our pitching staff giving up 81 more runs than we gave up last year," Hahn said. "Given that we have our opening day starter back in John Danks that Jake Peavy — knock on wood — should be putting together another healthy season; that Chris Sale and Jose Quintana are a year older; not to mention how much we like our bullpen. If everyone’s healthy, we don’t see ourselves giving up 81 more runs than we did last year."
PECOTA doesn't seem to expect the team's healthy pitcher trend — which coincides with the arrival of pitching coach Don Cooper in 2002 and the presence since 1979 of trainer Herm Schneider — to continue. No matter how professional their respective coaching and medical staffs, eventually, the law of averages will catch up to White Sox pitchers. Can the White Sox count on Peavy to be as healthy as Buehrle was, for example?
"If they’re right in terms of health projections, and all of a sudden there’s 150, 200 innings to fill, because of guys outside of the ones I mentioned in the rotation, then they might be right," Hahn said, speaking of Gavin Floyd (and ... Hector Santiago
"But then, historically — knock on wood because we don’t like going down this path — we’ve been able to keep guys healthier and get more out of our starters than they projected."
Hahn brought up the bullpen as an X-factor again.
"I think our bullpen also tends to be a little stronger than they give us credit for," Hahn said. "And again, predicting reliever performance is a difficult thing."
Sometimes in the projection business, there's just a hard-to-explain outlier. For PECOTA, it seems to be those pesky, perky White Sox pitchers.