Quantifying human movement

A missile-tracking technology re-appropriated to tracking player movement on a sporting field is one of the most revolutionary developments in sports analysis in my lifetime. Stats LLC's SportVU product uses multiple cameras to track player movement on a field of play. The possibilities for a deeper understanding of any game are mind-boggling.

Stats LLC gave Zach Lowe of Grantland.com some SportsVU data from the 2012-13 NBA season, and he posted some observations from his dissection of that dataset. It's a goldmine. What's most shocking is that only 15 of the NBA's 30 teams have installed the SportVU camera system in their stadiums.

Basketball can be thought of a network in which you have 5 nodes distributed through space, and the goal is to get the ball to the node with the greatest chance of scoring a basket (which in turn can be based on field goal percentage space charts). Spatial analysis can analyze how consistent a team is at getting the ball to the proper node, regardless of outcome.

I first heard of the practice of tracking players' movement through space from baseball, where it was used to track defense. Traditional defensive stats are a very weak measure of defense because a player like Derek Jeter with poor lateral movement might not reach a ball that another player with greater range might get to but mishandle for an error.

Sportvision's FIELDf/x service uses cameras to track the position of players at the moment the baseball is hit, and it also tracks the trajectory and velocity of the batted ball. It allows front offices to more accurately assess the defensive skill of a player, distinguishing between a play that looks difficult and a play that actually is difficult. Gold Gloves in baseball are still the result of a flawed voting system, but thanks to the new defensive tracking tools, we have a much better idea who the real defensive stars are at each position. The only question now is whether the data will be made available to the public, or whether it's more lucrative to keep it private, licensable only by MLB teams.

Such technology is being used in multiple sports, now, and the result is an increase in the accuracy of measuring a person's true output or performance. This should, in theory, lead to more equitable salary allocation.

One of the most unique things about sports, in contrast to the business world, is that salaries are largely public. Technology companies often speak of transparency as a core principle, yet the one area transparency has not worked is in compensation (except for really senior officers at publicly traded companies).

The primary reason it wouldn't work in business is that measuring performance is nowhere near as simple as it is in sports, where all players perform in front of an audience of millions in every contest, and where widely available statistical output is seen as a decent proxy for value.

In fact, in business, measuring skill and value is so difficult that even interviewing and hiring are processes with high error rates. Separating a person's value from their context is extremely complicated, and even in sports, statistics are a poor measure of future production when the competition is unevenly distributed, as it is in high school, college, and the minor leagues (if it were easier to evaluate athletes in those levels, you'd expect professional production of draft picks to follow a nicely descending curve down and to the right, but instead it's quite lumpy).

For certain crafts, it's become easier to evaluate skill. For designers and programmers, for example, you can examine a body of work, say a portfolio or previous code, and you can ask candidates to solve problems live. Given the money involved in sports, I'd expect more and more to be spent on trying to evaluate a player's true performance. That might involve asking potential draft picks to wear tracking devices during games, or teams might simply start installing systems like SportsVU in more and more stadiums so player performance can be more accurately quantified.

As a sports fan, I'm curious to see what types of data we'll be able to see in broadcasts of the future. The number of miles run by each soccer player on the field. The average angular differential between where a tennis player hit a ball and where their opponent was at the moment of impact. Which center fielders in baseball would have likely caught a ball that ended up splitting the gap for a double against the player in center field at that moment. The open shot percentage that any NBA team gives up, on average (an open shot being defined by the nearest defender being over X feet away at the time of the shot). The possibilities are endless.

Some find the increased statistical dissection of sports dull and soul-deadening, but I suspect the opposite is true, that when we translate human motion into figures, we'll truly appreciate just how remarkable the world's greatest athletes really are.