What advanced tracking data reveals about NBA shooters

AP Photo/Sue Ogrocki

Who is the best shooter in the NBA? Stephen Curry is a name that seems to be popping up a lot lately. But how can we know for sure? If we use field goal percentage, the classic way of measuring shooting that everyone seems comfortable with, Steph (50.4) is worse than his backup Shaun Livingston (53.2). Steve Kerr must know something we don't.

The biggest advance still trying to gain acceptance is effective field goal percentage (EFG), which gives one and a half times credit for a made 3-pointer because apparently three is 1 1/2 times 2. Using EFG, Steph (62.8) is better than Shaun (54.2), but he is still worse than Andrew Bogut (63.1). So Steph is not even the best shooter on his own team. But our eyes know that what Steph does is way harder than what Andrew Bogut does. Bogut probably knows this as well.

Defining quantified shot quality (qSQ)

Most fans have an intuitive sense of how hard any shot is, but we have never been able to quantify what we could see until now. The NBA has been tracking locations of the ball and the players (on the court) league-wide since the 2013-14 season. This has created a gargantuan pile of data that in its raw form is essentially useless to most people. But, if taught properly, this data can allow a machine to understand the game and show us amazing things we previously never could see.

One of those amazing things is the shot quality of every shot taken in NBA -- except instead of an intuitive sense, we get a number. How does a machine learn that? It can look at every shot in the NBA, consider the positions, movement, distances, and velocities of the shooter, the closest defender, the next closest defender and the ball (and much more) to make a prediction of the likelihood of that shot going in, if it didn't know who was shooting it. This prediction is what we call the quantified shot quality (qSQ) of that field goal attempt. Another way to look at qSQ is that it's the effective field goal percentage (EFG) of that particular shot for an "average" NBA player.

Let's look at some examples and see if it stands up to scrutiny. For the 2015-16 regular season, we looked at 173 players who took 500+ shots with the associated tracking data. Here are the five with the highest and lowest qSQ.

For the players with the highest shot quality, we have three players who take almost all their shots in the paint with the vast majority being catch-and-shoot (DeAndre Jordan, Dwight Howard, Kenneth Faried) and two players who take almost all their shots in the paint or from 3-point range where most threes are catch-and-shoot (Kent Bazemore, Draymond Green). These are all generally considered high quality shots. On the other side, we have four players well known for midrange off-the-dribble shots (Kobe Bryant, DeMar DeRozan, Carmelo Anthony, Jamal Crawford) and a big man who takes half his shots from midrange (Marc Gasol).

The qSQ of a shot reflects what happened, but not why. Players may end up taking easier or tougher shots because of their team's scheme, the player's decision-making, or because they have to bail their team out of a failed play. But, qSQ does represent the quality of the resulting shot.

Defining quantified shooter impact (qSI)

Evaluating shooting by whether the shot went in or not mixes up two things: the quality of the shot, and the impact of the shooter. We didn't have a way of separating the two until now. If we take a player's shooting performance, measured by EFG, and subtract shot quality, measured by qSQ, we have a number that tells us how much better than an "average" NBA player the shooter is for the shots they take. We call this quantified shooter impact (qSI).

If we evaluate shooters by EFG alone, many of the highest-rated players will be ones who get easy shots (and it will look like Curry and Bogut are roughly similar). But by using qSI, with players who took 500-plus shots in 2015-16, we have the following top five:

That list looks pretty good. Steph is at the top. J.J. Redick quietly had a ridiculously great season. Kevin Durant, Kawhi Leonard and Karl-Anthony Towns have also been popping up as being quite good at basketball. That's also a complete starting five. If we used the top five using EFG, we'd have three centers. But that doesn't mean centers can't also have great impact.

Over the past two years, 170 players took at least 1,000 shots. Of those, players DeAndre Jordan and Hassan Whiteside were No. 1 and No. 12 in qSQ, meaning they had the easiest and 12th easiest shots in the league. They were also No. 8 (Whiteside) and No. 10 (Jordan) in the league in qSI. That means they added significantly more impact than the vast majority of the league, even though their shots were among the easiest in the league.

The qSI metric can also shine the light on greatness that we see when watching the game. Over the past two years, two certain Hall-of-Famers, Chris Paul and Dirk Nowitzki, ranked No. 23 and No. 72 in EFG, respectively. But they also happened to be taking some of the toughest shots in the league. Paul was No. 153 and Nowitzki was No. 160 in quantified shot quality (qSQ). Taking this into account and looking at quantified Shooter Impact (qSI), Paul was No. 6 and Nowitzki was No. 12. Yes, they are as great as you think they are.

This reflects what is great about machines understanding sports. It's not that they take over and create things that only machines can understand. In fact, it's the opposite. People and machines can work together to measure things in ways that more accurately reflect what we see with our eyes. And instead of just relying on an intuitive eye test, which isn't always right, we can quantify our guesses and see how they measure up. Does DeAndre get easy shots? Yes, No.1. Is he still ridiculous at finishing? Yes, No. 10. Does Dirk take tough shots? Yes, No. 11. Is he ridiculous at shooting? Yes, No. 12. Is Steph the best? Yes, No. 1. And, we are just scratching the surface of what is possible.