Atlanta Braves PECOTA 2020 vs ZiPS: Just educated guesses
By Fred Owens
Lowered expectations
PECOTA uses its projections for individual players to generate a picture of a team’s performance. Those projections see wins decreasing ($) because of roster changes and regression.
PECOTA predicts the offense will hold its own, but it doesn’t expect our starting pitchers to perform as well.
". . . Ronald Acuña, Ozzie Albies, and Mike Soroka is returning and they’re all 23 or younger. Freddie Freeman’s been a four-win player for four straight years. . .the Braves (signed) free agents (that) PECOTA sees them adding 6.5 wins. But they also lost (players) who contributed 9.2 wins. . .nearly a three-win deficit. . . PECOTA expects the Acuña/Albies/Soroka/Freeman core to generate nearly six fewer wins than in 2019. The bullpen’s deep, but the rotation falls off after Soroka and Hamels.(My emphasis added)"
PECOTA doesn’t show any starter ($) averaging six innings a start this season, though Mike Soroka is close at 5.9, so I’ll round that to six too.
If you follow the link, you’ll see that between Mark Melancon (40), Will Smith (45), and Shane Greene (15), PECOTA projects a total of 95 saves. The algorithm doesn’t consider the change of teams or the way manager Brain Snitker will use his pitchers in 2020; the projection assumes similar use to last season.
Everyday-player starts by position are similarly skewed, and have little relationship to the way the team will line up.
Dan Szymborski developed ZiPS around the same time Nate Silver wrote the first PECOTA. Like PECOTA, ZiPS is a computer projection system that uses multi-year statistics, with more recent seasons weighted more heavily.
"When estimating a player’s future production, ZiPS compares their baseline performance, both in quality and shape, to the baseline of every player in its database at every point in their career. This database consists of every major leaguer since the Deadball era . . . and every minor league translation since . . . the late 1960s. . . . ZiPS then generates a probable aging curve — both midpoint projections and range — on the fly for each player."
PECOTA also uses this approach, so how do the projections match up?