Last December, Dave Cameron discussed the mistake of analyzing a team’s chances in the coming season by simply adding and subtracting the performances of players acquired and lost during the offseason to the remaining players’ performance the previous season. For example, if a team hit well except for one spot in their lineup, and they fill that hole with a better player in the off-season, some are tempted to say that one can simply add the new player’s previous performance to the rest of the team’s, and the lineup is now great. Dave pointed out that this is a mistake, because it doesn’t take into account the way in which individual players are likely to age, regress to the mean, perform in a new context (in the case of players changing teams), and so on. In the spirit of Dave’s reminder that the previous season is not a “constant” to be used in analysis of the next, we should keep in mind that current season statistics aren’t a constant when handicapping the playoffs.
Given roughly the same playing time in 2010 as in 2009, a player’s 2010 performance is obviously more relevant. But just as we shouldn’t expect a player to repeat his 2010 performance in 2011, we shouldn’t expect him to duplicate his 2010 performance in the 2010 playoffs. Regression to the mean isn’t a process a player goes through over the winter, but is an essential part of how a player’s “true talent” is estimated at any point in time. When Zack Greinke was not projected to repeat 2009’s 2.16 ERA in 2010, it wasn’t on the basis of him becoming less talented, but rather that he was quite unlikely to have been a true talent 2.16 ERA pitcher in the first place — he was an very talented player who nonetheless was likely pitching “over his head” during 2009. Ryan Hanigan has hit well this season, especially for a catcher, but he’s probably not going to hit for a .368 wOBA in 2011. For the same reasons (past performance, regression to the mean, etc.), he probably isn’t a .368 wOBA true-talent hitter right now, either.
Another mistake is analyzing playoff match-ups using a team’s corporate statistics from the current season, whether it is team OPS, ERA, wOBA, FIP, WAR, or whatever. In addition to the way in which this neglects the difference between observed performance and true talent on an individual level noted above, it also passes over the practical differences between how playing time works out in the regular season as opposed to the playoffs. One obvious example: the Twins offense was 78 park-adjusted batting runs above average this season, good for third best in baseball. However, Justin Morneau accounted for about 36 of those before his concussion — almost half, and Morneau is out for the playoffs. Javier Vazquez pitched 144 (mostly terrible) innings for the Yankees during the regular season, which obviously impacted their team stats, but he won’t being seeing that proportion of action during the playoffs. The Giants team batting runs include a -7 from Bengie Molina, who won’t be playing in the playoffs — not for San Francisco, at least. The Phillies batting statistics reflect Chase Utley only playing 115 games this season; I’m guessing he’ll start a bit more than 70 percent of the Phillies’ playoff games.
One can think of numerous similar cases. Moreover, team season stats include playing time for bench players to rest starters — something that rarely happens in the playoffs. This all may seem rather obvious, but when using raw team statistics to analyze playoff match-ups, these are the sorts of things that are forgotten.
It is understandable why people use team statistics from the current season to dissect post-season match-ups, and I’ve been guilty of doing it myself. But when doing so, we’re using a meat cleaver rather than a scalpel.