Back in the 1980's, long after the Dark Ages but before the Digital Age, winter was a calmer time for the serious baseball fan. You waited for the newspaper for news of a major trade. You waited for Topps to release the new baseball cards so that you could have a new season of stats for your favorite players. You curled up by the fire and read a book that you checked out of the library.
As the calendar turned to February, you eagerly anticipated the most important annual publication of the year. A publication that you did not need to read cover to cover, but could bounce around to find the most engrossing takes. Not a glossy magazine, but a thick tome called the Bill James Baseball Abstract.
Bill James' genius was more in asking the right questions and writing elegant prose than in complex statistical analysis. One concept he developed was similarity scores, originally for Hall of Fame debates. This seeks to find players from the past who are most like the player in question. You can see this at baseball-reference.com today if you scroll far enough down a player's page. For instance, the players with a career most similar Jack Morris were Dennis Martinez, Andy Pettitte, Bob Gibson, Luis Tiant, and Jamie Moyer. Cherry-picking Bob Gibson advances Morris' argument, but also shows the limitations of this approach with exceptional players.
The similarity scores concept was extended to predicting the next season of an active player, and became PECOTA. Rob discussed this last week when presenting the PECOTA projections for the American League Central Division.
I examined the ZiPS projections for the Tigers a few weeks ago. ZiPS uses the past three or four years of data to project the next season using weighted averages. PECOTA instead seeks to understand the career arc for the most similar players. PECOTA thus can produce significantly different projections for a given player. For instance, PECOTA sees Rick Porcello breaking out to a 3.5 WAR in his age 26 season, which would be his best season yet. Such improvement is hard to predict if simply averaging his past four years.
While PECOTA is optimistic about Porcello, it is not so much about most of the Tigers. It is as if there is some residual Bill James inspired Boston bias in the projections.
PECOTA is not so fond of Avila's on-base percentage, and sees little point to arguing whether McCann or Holaday should share catching duties.
Miguel Cabrera will be good. You know that whether you look at his statistics the past few years, or knowing that the most similar batters through age 31 are Hank Aaron, Frank Robinson, Mel Ott, Ken Griffey, and Albert Pujols.
Ian Kinsler's four wins are encouraging, as is Jose Iglesias' projection of 1.3 WARP when considering that defense is hard to quantify.
Nick Castellanos must have some scary comps, because ZiPS is more optimistic than PECOTA.
PECOTA does not like the outfield. J.D. Martinez's 2014 season will fade quickly, as he returns to replacement-level performance. Anthony Gose will disappoint, failing to reach base 30% of the time and thus limiting opportunities to use his speed. Rajai Davis will outperform Gose and cause debates about who should start in center field. And Yoenis Cespedes is merely an average player. I assume the algorithm could not find sufficient similar Cubans to make a reasonable forecast.
Victor Martinez will have a batting average over .300, but we know that without fancy algorithms.
The starting pitching projections are really making me scratch my head. David Price has the same value as Rick Porcello. Justin Verlander will put up identical numbers to Price. Anibal Sanchez will not miss a start, but will fail to impress. Alfredo Simon will not be worthy of a spot in the rotation, but will still make nearly every start. Shane Greene will struggle through 23 starts with a 5.60 ERA and 1.62 WHIP before being replaced by an only slightly more effective Kyle Lobstein.
But the weak rotation will be supported by a decent bullpen. Joe Nathan and Joakim Soria have nearly identical projections with 9.17 strikeouts per nine innings and a 1.2 WHIP. Al Alburquerque and Bruce Rondon will be reasonable options as long as first base is open, walking a batter more often than once every two innings. Ian Krol will fail to secure the lefty spot in the ‘pen, but Tom Gorzelanny will be solid.
The financial analysts are required to chant "past performance is not necessarily indicative of future results". Perhaps the 2015 Tigers can teach PECOTA this life lesson.