MLB Betting: How Accurate is Nate Silver's PECOTA System?

Jason Lake

Thursday, July 16, 2015 12:08 AM UTC

Thursday, Jul. 16, 2015 12:08 AM UTC

He's made quite the splash since his days with Baseball Prospectus. Nate Silver is all over the media landscape now, but he's still helping us with our baseball picks.

Jason's 2015 record as of July 15: 31-24, plus-7.36 units ML; 0-3, minus-3.30 units Total

If you had to name one stat guru, you'd probably come up with Bill James. Then it would be Nate Silver. Both gentlemen ended up on the TIME 100 list – James in 2006, and Silver in 2009. The latter has become an even bigger name these days thanks to his work in political forecasting over at FiveThirtyEight, but it was baseball where Silver first made waves. Now that we're at the halfway point of the 2015 season, let's take a moment to look at Silver's top contribution to our MLB picks, the PECOTA system.


Down in the Silvermine
Back at the turn of the millennium, Silver got his first “real” job as an economic consultant with KPMG in Chicago. Not surprisingly, he hated it. So Silver, a baseball nut, spent his time working on a stats-based player projection system, which he sold to Baseball Prospectus in 2003. Silver got a piece of the company in exchange, and the rest of the world got introduced to PECOTA.

So what is it? PECOTA is a proprietary system that uses similarity scores to predict how a player will perform, based on how the other 20,000-odd MLB players have done over the years. The name stands for Player Empirical Comparison and Optimization Test Algorithm, but really, it's a nod to former MLB infielder Bill Pecota. The “backronym” has inspired similar metrics in other sports, including Kevin Pelton's SCHOENE forecasts for basketball.

PECOTA was originally targeted at fantasy baseball, and it did very well at projecting things like OPS and ERA. But as time went on, it was also fitted to project team performance as well as individual performance. There was a time in the mid-2000s when PECOTA was becoming increasingly accurate with its projections, although things went a bit haywire in 2008. BP has made some changes to the formula since then, although Silver's watch ended in 2009.


Let's Play 162
As one of the first Moneyball-era metrics, PECOTA still gets quoted in the press every spring, although you need to subscribe to BP to get all the stats. Many teams seem to be at or near their projected levels thus far, which is to be expected in baseball – MLB teams tend to finish around .500 more than the other major pro sports. Playing 162 games instead of 16 will do that.

Among the most glaring errors at this point: the Boston Red Sox (42-47, –11.50 units) were projected to go 87-75, and the Kansas City Royals (52-34, +16.85 units) were pegged at 72-90. But instead of throwing PECOTA out with the bath water, let's red-flag those teams as two potential candidates for some regression during the second half of the season. And let's thank Silver for not sticking with a job he hated.

comment here