Todd Whitehead

March 17, 2023

The old school position labels have a flaw. Go to Basketball-Reference and look up LeBron James, for example, and you’d find that during the last seven seasons he’s been listed variously as SF, PF, SF, PG, PG, C, and now PF again. And, yes, I realize that LeBron is just #BuiltDifferent but that’s literally every position possible, from 1 to 5. The problem with the conventional position labels is that they try to bundle together a player’s offensive role and his defensive responsibility into one package, which doesn’t work very well for describing a player like LeBron who is big enough to guard 6’9 opponents on defense but skilled enough to handle the ball on offense. Is he a PF or a PG? It’s hard to say.

That’s why our team has been working on making functional descriptions of how players play that decouple offensive role from defensive responsibility. Our offensive roles are defined by how a player tries to help his team score points, either through the shots that he takes himself or the ones he sets up for his teammates with his playmaking. Specifically, we want the Synergy offensive roles to reflect how often a player’s team asks him to do each of the following tasks:

• Ball handling
• Creating advantages
• Exploiting defensive rotations, ie. “keeping the advantage”
• Finishing (at the rim or from three)
• Spacing, eg. via off-ball screen setting, cutting, and gravity, and
• Positioning, ie. standing and catching the ball in various places around the court.

You can read all the nitty, gritty details about how we’re using Synergy data to define offensive roles in a separate post but, basically, we’re sorting players based on different combinations of their play-type and shot-type rates (%TIME stats) that best describe the ways in which they try to help their teams score.

Our goal was to create groups of players whose characteristics differ in ways that are meaningful to coaches and scouts in a practical sense and to have names for these groups that resonate with those coaches and scouts on an intuitive level. So, to make sure that our sorting process is working the way we intended, we validated our approach with publicly available player-tracking data from the NBA site. This was information which was *not* incorporated in the role sorting process, so it should be a fair way to check our work.

For example, a sharp coach might expect the players included in the ball handler groups to have more time of ball possession (a stat which is captured via player-tracking data) than the players included in the non-ball handler groups. For time of ball possession and each of the other tracking stats that are a direct analog of one of the six role-defining tasks – like paint touches, drives, etc. – we can use a “box and whisker” chart to check how well the sorting process was able to resolve the differences between the groups that were formed (using the previous NBA season as a test case).

In each row of each of the box and whisker charts that follows, half of the players’ dots are located inside the boxes, the lowest quarter of the player’s dots are past the left edge of the boxes, and the highest quarter of the player’s dots are to the right. The line through the middle shows the typical value (the median) for that offensive role. If the sorting process worked well and the groups are meaningfully separated along each important dimension, we will find small boxes (ie. there will be similarities within each group) and boxes which are not completely overlapping when we look down the page across the 11 groups (ie. there will be differences between the groups).

A player’s ball handling responsibility was one role-defining feature, which we can approximate using time of possession (via player tracking data).

As expected, we find that the three ball handler groups tend to have more time of possession (ie. more ball handling responsibility) than the wings or the bigs. Specifically, the 25th percentile time of possession in each of the ball handler groups is larger than the 75th percentile time of possession in each of the non-ball handler groups. In other words, the pink boxes do not overlap with the green and orange boxes. This is validation that – although we did not use player tracking data during the sorting process – we were able to effectively capture this characteristic role-defining feature of ball handlers.

Looking more closely, we can also see that among the four wing groups, the playmaking wings have the highest median time of possession. Likewise, among the four big groups, the playmaking bigs have the highest median time of possession. On the other hand, groups of “finishers” like the spot-up shooting wings and the rim-finishing bigs tend to be responsible for very little possession of the ball.

Another role-defining feature is court positioning, which we can approximate using paint touches (ie. the number of times a player caught the ball while standing in the paint, via player tracking).

As expected, we find that the bigs tend to have more paint touches than the wings or the ball handlers. Specifically, the 25th percentile paint touch rate in each of the big groups is larger than the 75th percentile time of possession in each of the non-big groups. In other words, the orange boxes do not overlap with the pink and green boxes. Again, this is validation that – although we did not use player tracking data during the sorting process – we were able to effectively capture this characteristic role-defining feature of bigs.

It’s worth pointing out that we’re using the term “big” here as a position (ie. a big is an interior player) and not as a physical description. Indeed, not all of the players who play the “big man” role on offense will actually be tall. Gary Payton II, for example, landed with the rim-finishing bigs for the 2021-22 season despite being only 6-foot-2. And, although he’s not physically big, Payton had a higher paint touch rate than all but two of the players in the wing or ball handler groups. Remember also that these are offensive role groups and defensive responsibilities will be characterized separately (Payton is a rim-finishing big on offense who mostly guards playmaking and scoring ball handlers on defense, for example).

Looking more closely at the paint touches chart, we can also see that among the four big-man groups, the rim-finishing bigs have the highest median paint touch rate and the stretch bigs have the lowest. This is an example of the sorting process working well, as we would expect to find the rim-finishing bigs hanging out around the rim and the stretch bigs to be stretching out further from the hoop. This is the type of 1:1 correspondence between the names of the offensive role groups and the characteristic features of their constituents that we hope will make the group labels resonate with coaches and scouts.

Corner three rate is another opportunity to validate the capacity of our role sorting process to capture where on the court a player is standing (and shooting).

Although we did not include shot locations as inputs in the sorting process, we find that the groups were ultimately separated by corner three rates anyways. In general, wings tend to shoot more of their shots from the corner than ball handlers or bigs (the four wing groups have higher median corner three rates than the rest of the groups). Among the four wing groups, spot-up wings get the biggest chunk of their shots in the corner. Among the three ball handler groups, secondary ball handlers take more corner threes than the other two groups. And, among the four big groups, stretch bigs take the highest rate of corner threes.

Another role-defining feature is a player’s ability to create advantages for his offense, which we can quantify using potential assists, ie. passes that were followed immediately (after less than 2 seconds) by a shot (whether it was made or missed) via player tracking.

For each of the bigger groupings – among ball handlers, wings, and bigs – the playmaking group is the one with the highest potential assist rates. Again, this indicates that, although we did not use player tracking data in the sorting process, we were able to produce groupings that reflect reality in a way that should resonate with coaches and scouts (ie. we believe they will hear “playmaking ___” and think of players who set up their teammates to score with passing and assists). On the other hand, groups of play finishers, like spot-up shooting wings and rim-finishing bigs, tend to create fewer advantages for their teammates (at least on the ball), hence they have lower potential assist rates.

We can also characterize a player’s ability to create advantages for his offense based on drive stats. Our loggers keep track of possession-ending drives and that information was incorporated in the role sorting process (ie. drives-to-hoop rate and drives-to-shot rate were both inputs we used), but this chart shows a slightly different measure of drives per team possession on the court via the player tracking data.

The drive chart reveals a similar pattern as the potential-assist chart (and, for that matter, the time of possession chart), ie. ball handlers tend to drive more than wings or bigs and playmaking wings and bigs tend to drive a bit more than the average wing or big. However, the additional distinction here (vs. those two previous charts) is the relatively higher levels of driving among scoring ball handlers and slashing wings. It’s also apparent that, while playmaking bigs do create advantages for teammates, they are less likely to do so with drives than playmaking ball handlers or wings.

For bigs, another way to assess advantage creation (as well as court location) is by counting post touches. Again, our loggers catalog end-of-possession post-ups in great detail, but – for the sake of objectivity – in the next chart we are using post touch rates derived from player tracking data (as an external check of our process).

Not surprisingly, post-up bigs tend to have the highest rates of post touches. In general, bigs tend to get more post touches than wings or ball handlers although there are some exceptions like LeBron James (a scoring ball handler) and Marcus Morris Sr. (a playmaking wing) who post up quite a bit. Among the bigs, rim-finishing bigs get the fewest post touches. In fact, a player’s post-up rate is what we’re using to separate a rim-finishing big (like Clint Capela) – who gets lots of paint touches but does not create advantages on his own – from a post-up big (like Joel Embiid) who can create his own advantages in the post.

We can characterize how often a player is asked to finish a play with a three by looking at his catch-and-shoot 3-point attempt rate. Here I’m using a touch time of 2 seconds or less (via player tracking) as a way to zero in on catch-and-shoot attempts, but the results are similar if we use the filter of no dribbles before the shot (or, equivalently, Synergy’s own catch-and-shoot tagging).

We did not include shot locations as direct inputs in the sorting process, but we see that the groups were split meaningfully by catch-and-shoot 3-point attempt rate. In general, wings tend to get more of their shots as catch-and-shoot threes than ball handlers or bigs. Among the four wing groups, spot-up shooting wings and dynamic shooting wings get the biggest chunk of their shots from catch-and-shoot threes. Among the three ball handler groups, secondary ball handlers take more catch-and-shoot threes than the other two groups. And, among the four big groups, stretch bigs take the highest rate of catch-and-shoot threes.

Finally, we can characterize how often a player is asked to finish plays at the rim by looking at his shots-at-the-rim rate.

As expected, the rim-finishing bigs get more of their shots at the rim than any other group. Notably, the distinction between a rim-finishing big and a stretch big is very obvious from this view. Among the four wing groups, slashing wings have the highest shots-at-the-rim rate and, correspondingly, the lowest catch-and-shoot three rate – which is what sets them apart from the other wing groups.

These eight box-and-whisker charts show that the role sorting process does an effective job of resolving the offensive role groups across at least four of the role-defining features: ball handling, creating advantages, finishing (at the rim and from three), and positioning. The remaining two offensive tasks of exploiting defensive rotations and spacing are more difficult to assess with publicly available player tracking stats but we can use stats that we derived ourselves from the Synergy logging as a final check. For example, here’s a look at how often players from each group tried to score by using an off-ball screen or handoff (on a z-score transformed scale that compares each player to the league average).

This is cheating a little because we used this movement shooting rate to help create the groups, so to circle back at the end and say, “See! The groups are sorted really nicely by the stats we used to sort them” is not the most convincing form of validation. Still, we can see that the process worked the way it was supposed to work, here, as the dynamic shooting wings (by definition) are the group with the highest movement shooting rates. Notice how the separation between groups is especially stark in this chart (ie. the dynamic shooting wing box doesn’t overlap with the other boxes at all). In the end, movement shooting rate is only a surrogate for the off-ball space creation but it’s not the role-defining feature itself. The same limitation would be true for a chart using any of our other player-sorting inputs, so we won’t bother going through the rest.

Finally, here is one counter example of a box-and-whisker plot for a player attribute that we decided *not* to treat as a role-defining feature, usage rate.

Because usage rate was not included as a direct input for the role sorting process, many of the groups’ boxes are relatively big and many of them overlap each other. In particular, the playmaking wings, slashing wings, dynamic shooting wings, playmaking bigs, and stretch bigs all have very similar median usage rates. Each group has some players who are high-usage players and some players who are low-usage players. Still, even though we put our emphasis on characterizing play styles and not on sorting players by their production, some latent patterns were revealed. For example, scoring ball handlers tended to have the highest usage rates and spot-up shooting wings and rim-finishing bigs tended to have the lowest.

Again, our goal was to make offensive roles that resonate with coaches and scouts on an intuitive level without any elaborate explanation. So, boiling down all the charts above, here is a short sentence to describe the characteristics of each group:

• Playmaking ball handlers have high on-ball scoring rates and high playmaking rates,
• Secondary ball handlers have high on-ball scoring rates with some spot-up, catch-and-shoot scoring,
• Scoring ball handlers have high on-ball scoring rates with very little spot-up, catch-and-shoot scoring,
• Playmaking wings set up their teammates to score more than most other wings.
• Slashing wings drive to the basket more than most other wings,
• Spot-up shooting wings have high rates of spot-up, catch-and-shoot scoring,
• Dynamic shooting wings try to score by using off-ball screens, handoffs, and dribble jumpers,
• Playmaking bigs set up their teammates to score more often than most other bigs,
• Post-up bigs have high rates of scoring or passing from the post,
• Stretch bigs have high pick-and-pop and catch-and-shoot rates, and
• Rim-finishing bigs try to score by rolling, cutting, and rebounding.

We’ll plan to include these brief descriptions of each role as tooltips on the team site for now and, hopefully, they will eventually become familiar terms. We’re hoping our new offensive role descriptions will help coaches prepare for and game plan against their opponents and help scouts to better understand the contexts in which prospects and recruits have played their basketball in the past and to imagine how those prospects and recruits might play differently in a new context. And we’ll have more to say about how, specifically, coaches and scouts might use these new metrics to develop winning team strategies in future posts, so check back here soon!

Todd is building tools to help coaches, scouts, and players find winning team strategies as part of Synergy’s Analytics and Insights Team. He creates inviting infographics, engaging charts and interactive displays that make data compelling and accessible. Follow him on Twitter @crumpledjumper