Post-Season Pre-Mortem

(This article can alternately be found at www.WriteOnSports.com)

Here are my postseason predictions: analysts will incorrectly pick a huge number of the series from the MLB playoffs. Even the best analysts get caught up in game-by-game momentum swings, tactical decisions, and individual trends that ought not be generalized over a longer period. For instance, the main articles on Baseball Prospectus on Thursday dealt with manager moves and play execution from Wednesday’s games, and the three-week strategic implications of postseason rosters- the most significant one being Joe Beimel’s injury-induced exclusion from the Dodgers’ postseason roster. Another of my favorite baseball sites, Baseball Think Factory, linked to a handful of knee-jerk reactions about one or two games, making big claims about teams’ directions and fundamental essences. Here is the bottom line: the postseason usually features a bunch of really good baseball that is really fun to watch, but it does not provide meaningful statistical data for analysis, especially on the fly.

While everyone knows that short series are unpredictable, I think many fans tend to forget just how short the postseason really is. Consider the total number of games for a moment. If every series goes the distance by today’s postseason structure, you will see eight teams combining to play a total of 41 games- not that many more than an individual starting pitcher starts over the course of the year. Even so, most series do not fulfill their maximum length potential. Setting the averages right in the middle- 4 games in the short series and 5.5 in the longer ones- only about 358 games have been played in the 11-year history of the current format, compared to 2430 total games played over the course of a single major league regular season. Starting in 1993 and going backward, if the Majors had played the LCS-World Series format since its birth, you would have to go back to 1868, roughly the birth of truly organized baseball- to reach a single season’s worth of baseball games. Of course, the LCS did not exist until a few decades ago. Therefore, there were more Major League games played in the 2006 regular season than in all postseasons up to and including this one combined. Any good statistician will tell you that single-season data is unreliable. Thus, looking at all postseason games- from the Black Sox, to the original Bronx Bombers, to Stengel’s Yankees, to the Big Red Machine, to Torre’s current iteration of the Bombers- gives us the same overall perspective as the one we get from looking back to sometime in June 2006.

We constantly make two critical mistakes in evaluating players in the postseason. First, we speak of individual postseason performance as if it is somehow easily known even though there exists less than a full season’s worth of data on even the most experienced players. Secondly, we foolishly separate regular season and postseason ability, as if they are two separate skill sets. Think about a player like Reggie Jackson, a true postseason stud who everyone recognizes as a prime-time performer. Still, Mr. October played few enough October games that all of his stats could be easily skewed by variance. We tell and re-tell stories of Jackson’s three homerun game in the World Series as if some mystical event or unique psychological ability gave him the strength to perform at a supernatural level for a short while. In reality, players have three homerun games every now and then, and power hitters are more likely to do it than slap hitters. Eventually, someone was bound to hit three homeruns in a crucial World Series game, Reggie Jackson happened to be that guy. Because of that and a handful of other really strong games, the equivalent of a several very good weeks makes Jackson look like a legend. Chris Shelton hit like the postseason version of Reggie Jackson for a couple of months this season, but he had a larger sample from which to draw by the end of the year, and ended up being replaced by Sean Casey. I do not mean to say that Chris Shelton was ever on Reggie Jackson’s level as a player, but a 40 or 50 good games does not a legend make, no matter when those games were played.

Part of the misunderstanding comes from the logical fallacy that says that postseason games really are unique; that it would be one thing for a player to have a couple of hot months, but doing it under the bright lights and intense scrutiny of the postseason is a meaningful sample rather than a truly random one. Just because the games are not sequential does not mean that they are not subject to variance. If we selected every player’s career stats from a certain set of days, such as the 12^th-15^th of every month excluding August, some players would probably have .400 career BAs, others would have absurdly high numbers of HRs, and others would struggle more than one might anticipate. If you took a sample of only Tuesday and Thursday games, there would be a similar result. The better players would rise to the top, and the worse ones would go to the bottom, but some would look out of place by nothing more than completely random variance. The postseason is a little difference due to the level of competition, but the batters and pitchers are all better, and are all presumably trying as hard as they can, so there is no discernable advantage.

The same sort of knee-jerk reaction applies to the amateurish calculations of momentum. As I watch game two of the Dodgers and Mets from my couch on Thursday night, I hear commentators dooming the Padres and Twins to an early death. While the Cardinals and A’s certainly hold a meaningful advantage. On the other hand, I have a strong feeling that the artificially constructed storylines would change drastically if either series reached a game 5. If one of the teams down 2-0 comes back to win the series, the losing team will be condemned as a fundamentally flawed squad that never had a chance to win the title- “They didn’t have the lineup to survive,” “Their starting pitching was too average.” But being fundamentally flawed in a week means they are fundamentally flawed today, and that storyline certainly does not have legs today. Same with the Yankees, who are supposedly in trouble after losing game two, and having to win two out of three against a team that has played sub-.500 ball since mid-season.

So before we start the 2006 prognostication season, let’s take a deep breath and step away for a moment. Any team could beat any other team, and we have very little ability to predict the minute changes that really do have an influence from game to game. So let’s stop falling over ourselves talking about a single play or the 25^th spot on the roster, and enjoy the baseball.