Foul Territory

There's a revolution afoot in the world of sports analysis. It's been going on for quite a while now in the world of baseball, but the other major sports are catching up fast. For years, there were two ways to evaluate players. First, there were the conventional counting and rate stats (an example of a counting stat is RBI, while batting average is a rate stat). Second, there was what your eyes told you. Who's a good athlete, who's slow, who has a nice swing, or a good looking gait. That was all that was available for the longest time, and professional sports teams suffered for it, at least in terms of getting the most value for their money.

Bill James was one of the first to realize these inadequacies, for baseball at least, and work on trying to quantify value in a more accurate manner. Hundreds of others followed in his footsteps, and many more continue to develop their own ideas about performance analysis in baseball and other sports. The web site Football Outsiders has attempted to take a new look at professional football that goes beyond yards and points. They have developed some rudimentary methods for looking at the contributions of single players. One of these is called Defense-Adjusted Points Above Replacement. In basketball, people like John Hollinger have developed ways to better evaluate a player's contribution, equalizing players who get different amounts of playing time, and also determining how much to weight each statistic (e.g. rebounds, blocks, assists). John has developed a stat called Player Efficiency Rating that tries to assign each player a number that can be used to rank him against other players.

The theme tying all three of those sports together is the goal of isolating individual players from their teams in order to better understand individual value. Each of these sports is a team game, and the success of the team is a function of the individual contributions. What's difficult is determining what those individual contributions may be. Baseball got the jump on this one, because much of player evaluation in baseball is tied to the singular act of hitting a ball thrown by a pitcher. Every outcome is discrete and measurable, so determining value is not as difficult, given the right statistical tools. The argument tends to lie in which part of hitting is more valuable. Joe Morgan and Tim McCarver make it sound like bunting and stealing are way more important than they prove to be, while more sophisticated methods show that slow players who hit doubles and home runs and take walks have more offensive value than fast guys who are good bunters. For pitching, things like ERA are heavily influenced by the quality of the fielders behind you, and to add to the confusion, the batting average allowed by a pitcher on hit balls is at least partially random. Therefore, it seems that a pitcher who doesn't let the ball go in play is a valuable asset. Defense is the next frontier of performance analysis in baseball. In defense, there is much more interaction between teammates, so isolating individual value is difficult. It's a lot easier to play outfield next to Torii Hunter than it is next to Craig Biggio. In the same sense, a fast shortstop will take chances away from his third baseman. Gold Glove awards and fielding percentage are nearly useless, as Gold Gloves can be won from a single SportsCenter clip (Derek Jeter, 2004) and high fielding percentage only counts the balls that the player gets to, not the ones that he can't reach because he gets terrible jumps (Jeter again, coincidentally). Statistics like Zone Rating and Fielding Runs Above Average do a lot to assign value to individual players, but even the inventors of those stats realize their limitations.

The interaction between teammates makes evaluation in basketball and football more difficult. The more players on the field or court, the more difficult it is to isolate an individual contribution. Basketball is easier because there are less players and every player generates statistics, and many of those are somewhat independent of their teammates. That way, John Hollinger can come up with his player efficiency ratings based on things like points per shot and turnovers, because those things are easy to measure. As with baseball, measuring defense is not as easy, as steals don't tell the whole story, as a gambling player can get a lot of steals, but also get beat for a lot of easy layups. Team defense is also a hard thing to evaluate, but I know smart people are working on it. As for football, we've only reached the tip of the iceberg. There are 22 guys on the field at a time, and many of them never touch the ball, even if they're vitally involved with the offense. It's impossible to evaluate a running back without his blockers, receivers without the quarterback, and a quarterback without all three. Aaron Schatz over at Football Outsiders is doing some revolutionary work, but he realizes how far he is from understanding offensive and defensive line play apart from some rudimentary statistics.

It's this kind of thinking that is getting Ivy League graduates with little athletic experience but lots of analytical ability jobs with professional teams and allowing others to make a living writing for widely read sport-specific websites. For many, the goal is to see how to improve a team by helping that team determine where allocate finite resources to derive the greatest benefit. Figure out what is undervalued, and get it at a bargain. It's not a lot different than analyzing stocks, except that previous results do tend to predict future returns. Other sports will soon get on the performance analysis bandwagon once some smart person determines how to isolate individual performance.