Failed Sloan Presentations: Towards A Comprehensive Valuing Metric for Total Player Success

stats cam 1

Every new statistical revolution in the NBA has held forth the tantalizing promise of showing us the game more clearly, of allowing us to finally understand definitively how each player contributes to a team winning or losing. Whether it was something seemingly simple like keeping track of blocks starting with the 1973-74 season or looking at per 36 minute numbers instead of per game numbers or John Hollinger’s more complex player efficiency rating (PER) or the still nascent expected possession value (EPV), new stats and analytics have provided us with ever-greater fidelity, but total understanding has eluded us.

Until now.

As you may be aware, every team in the NBA now has a SportVU system installed in their arena. This system of cameras records everything on the court 25 times a second, providing “a breakdown of speed, distance, player separation and ball possession data,” according to SportVU’s website. Data from SportVU has already helped to separate the act of rebounding into component parts, to track where an optimal, “ghost” defender would move on a play and to let us know precisely how far a player runs in a game.

But the system doesn’t only track players; it follows the ball as well. This means it is now possible to track where the ball is on the floor at all times and — most importantly for us — to measure and record every instance of a player taking a shot and whether he missed it or made it. This is the event at the heart of this new metric.

Working backwards from the mountain of data that the SportVU system provides, it’s possible to break down the court into different regions that are more or less valuable, based on the volume of shots and shot difficulty, as calculated by some pretty complex formulas. By understanding where a player is shooting from, it’s possible to assign an abstract value to a successful shot. We’ll refer to this value as a Persistent Offensive Instance of Non-Trivial Shooting (POINTS).


The bulk of these shots (about 61.9 per game for the 2013-14 season) are taken from the area directly around the basket and extending out in a rough arc to around 23.75 feet (not a completely precise measurement, as you’ll see in a moment). For successful field goal attempts within this zone, we assign two POINTS to the player who made the shot.

Now just beyond this range of 23.75 feet (and slightly less around the edges of the court near the baseline — about 22 feet), our data indicates that made shots should be worth more. Some data points to a zone where a made field goal would be worth four POINTS, but that seems a little excessive so instead we’ll stick with assigning a value of three POINTS to made shots in this area.

One odd wrinkle we’ve found is that while the majority of shots (83.2 per game) are taken with all the players in motion (with a few notable exceptions, including James Harden on defense, for some reason), there are about 23.6 shots per game that are taken from an area near the middle of the widthwise axis of the court and 15 feet from the basket where the other players on the court line up on either side of a rectangular area around the hoop. (Sometimes the other players stand around the halfcourt line — it’s beyond the purview of this study to determine why.) These shots are of a much lesser degree of difficulty for players than most of those taken on the court (with a few notable exceptions, including DeAndre Jordan, for some reason), so for these shots we assigned a value of just one of the POINTS we are tracking.


By themselves, any one of these instances we’re tracking is nearly meaningless. Some players score a lot of these POINTS and others score fewer, yet a player on one team can score a lot and lose while a player on another team can hardly score at all while his team wins.

This is where it gets exciting: If you aggregate all this data across all the players for a given team on a given night and compare it to that same aggregation for all the players on the opposing team, the results are very highly correlated with which team wins the game. Very highly. Like nearly 100%.

By tracking these POINTS for the teams as a whole, you can quite easily see who won the game and with far greater accuracy than would come with only watching it without tracking these POINTS. As we all know, the “eye test” can only take you so far, and neither straightforward data like rebounds, turnovers and fast break points, nor more advanced measures like plus/minus or PER can give you the whole picture at a glance, but with POINTS, it’s simple. Look for example at what POINTS can show us about LeBron’s stunning performance against the Bobcats:


(Note about the diagram: The team that won is in red.)

This data is also expandable and modular. You can drill down, for example, and see that Miami as a whole scored more POINTS in every quarter except the fourth, especially in the third, where James notched a very impressive 25 POINTS en route to a total of 61 POINTS.  This isn’t just the tip of the iceberg. IT’S THE WHOLE ICEBERG.


Perhaps most excitingly, this metric — with some tinkering — can be applied in retrospect to historical NBA games to give us a better understanding of why a team won this game or that game. Of course, we’re months or maybe even years away from that, but the important thing is that this new metric brings us one step closer to understanding basketball in a total and comprehensive way.

Here’s to a bright tomorrow.


Follow Steve on Twitter @steventurous


Leave a Reply