The New NBA Heliocentrism: How teams revolve more around todays stars than they ever have before

Only a season and a quarter into their NBA careers, it’s inevitable that Luka Doncic and Trae Young will be forever linked by the draft-day swap that landed the former in Dallas the latter in Atlanta. It’s not simply the trade itself or Doncic’s emergence as a plausible MVP contender in his second year for the Mavericks, but also because of the similarity of their styles and how it might be a harbinger of a wider adoption of a new role of NBA stardom. As much as the conversation of the 2018 Draft will revolve around these two, their respective teams revolve almost completely around the young stars in a way that was was unprecedented but is becoming extremely common.

Advertisement

Star players have always been described in ways that expound on their centrality and irreplaceability to their team’s success, so to illustrate why this development of stars becoming stars in the more Copernican sense matters, quantifying that centrality is a necessary precursor.

Overall offensive contributions

That accounting for a player’s role is important in measuring his performance is not a new concept nor is it limited to the realm of statistics. Discussion of players has always included consideration of environment. We make allowances for a rookie forced to take on too large an offensive load too soon, or when injuries force a wing to assume point guard-like duties as a primary ballhandler.

But defining those roles can be a challenge. Various approaches have attempted to classify players, redefining the traditional positions into roles (here is an example using a common technique called k-means clustering), and these have produced interesting results, but have never really gained much currency because the “clusters” of players produced are not exactly self-explanatory and often require significant explanation. Plus, there are frequent situations where two players are “neighbors” but happen to belong to different clusters, even though they might have more in common with each other than with many members of their cluster. And if that sounds slightly confusing, that’s probably a more succinct way of demonstrating why these efforts have remained more in the realm of interesting curiosities than practical usefulness.

As is frequently the case, I think we can discuss roles in simpler terms instead of trying to solve everything at once. Instead of trying to label groups of players, start by more accurately describing what it is they do. For any number of reasons, this is easier when discussing offense. The most prevalent stat for discussing offensive role has been usage. At least, one version of usage. But we can pretty easily do better. Players do much more on offense than score. They take care of the ball, they set up teammates, they do interesting things off the ball, like cut and set screens. And yes, they also shoot to score.

Advertisement

There are metrics that purport to address some of these. Turnover% and Assist% are readily available on both NBA.com and Basketball-Reference. But these are, to my mind, two of the more misleading stats in common parlance. The thought behind them is solid. As noted earlier, we want to know how well a player takes care of the ball and how frequently he sets up teammates. However, due to the way these stats are calculated, they both, somewhat perversely, reward a player for shooting almost as much as for doing the thing the stat purports to measure.

Take Turnover%. The calculation is simply:

TURNOVERS / (TURNOVERS + FIELD GOAL ATTEMPTS + 0.44 * FREE THROW ATTEMPTS)

That means the quickest way to appear to turn the ball over less is … to jack up shots. Passing can only end badly from this standpoint as either it’s a nothing or it’s a turnover. So, fire away! Similarly, Assist% is calculated as the proportion of teammates’ made field goals while the player is on the floor. A player who carries a 30+% usage will necessarily reduce the denominator in that equation, making a given assist rate appear more impressive on the stat. So, I prefer to toss those stats out and start anew.

Starting with 2013-14, we have the tools to catalog many of those contributions. Off ball is still a challenge, at least with publicly available data, but the others are still approachable.

I originally called this approach “True Usage” but have settled on “Total Usage” as a better name, and it is made of three components.

• Scoring Usage: True shooting attempts (defined as FGA + 0.44 FTA) divided by total chances for which the player in on the floor. Over time, 0.44 has been shown to be a reasonable approximation of shooting possessions accounting for three-shot fouls, and ones and technicals, so it has become the convention in calculating both traditional usage and True Shooting Percentage stats. Even though a more precise count can be derived from play-by-play, the results are close enough for the ease of working with aggregate level stats to be perfectly acceptable.

Advertisement

• Playmaking Usage: Potential Assists + Free Throw Assists divided by total chances.  Potential assists and free throw assists are tracking data derived counts with fairly strict definition — passes from the play that saw the receiving player shoot or be fouled in the act of shooting in two seconds or fewer and nor more than one dribble after receiving the ball. This measure is probably overly stingy in a few ways, both directly in that we would recognize certain occurrences outside of those narrow parameters to be assists, and also “hockey assist” and further downstream situations where the player most responsible for the scoring chance is two or more passes removed from the shot. But I’m prepared to live with these drawbacks to remove the inherent bias of scorekeeper awarded assists while also giving the playmaker credit for setting up teammates, regardless of those teammates’ ability to make shots. Especially in this era of 3-point dominance, I think a player should get credit for finding a teammate open in the corner, even if that teammate is going to miss more often than he will make the ensuing shot.

• Turnover Usage: Turnovers divided by total chances.

These are summed to one number, so a player with a 25 percent scoring usage, 15 percent playmaking usage and 2.5 percent turnover usage would have a 42.5 percent total usage. Which would be among the highest rates in the league.

This also allows for a better measure of turnover propensity relative to offensive involvement. Defining True Turnover Percentage as Turnover Usage/Total Usage,  players can lower their “true turnover percentage” by playmaking as well as by shooting.

Since this discussion started with the Mavericks and Hawks, the chart below illustrates the scoring and playmaking usages of their respective rotation players:

This shows the degree to which the Hawks and Mavs do, in fact, revolve around their second-year players. But to some degree, the NBA has always been a stars’ league in this way. But to this extent? Not really.

New Style, New Analysis?

At least since Russell Westbrook averaged a triple-double for the season for the first time, we’ve been asking, “What does it all mean?”

Advertisement

Not in an existential sense, but more specifically in terms of basketball value. Even then there was a recognition that some of Westbrook’s statistical achievements were, at minimum, facilitated by the post-Kevin Durant eco-system of the Thunder. Those OKC teams had (largely successfully) been built with two dynamic, creative fulcrums surrounded by a series of rangy, defense-first athletes with more limited ball skills. And when Durant departed via free agency, the Thunder leaned in and tried the same model, only with a single engine. It mostly worked, with OKC averaging 48 wins per year across the three post-KD seasons.

An interesting outgrowth of the Thunder becoming so heliocentric is that Westbrook in effect broke certain statistical models. His 2016-17 season is rated as the best ever by Box Plus/Minus. As great as the accomplishment of averaging a triple-double might have been, it is hard to square “best season ever” with a guy putting up roughly league-average efficiency numbers for a 47-win team. BPM is built on a solid evaluation of historical trends, but it wasn’t designed or prepared to adjust for a player not playing by the rules in terms of things like hoovering up every available rebound off of a missed FT.

It turns out that the value BPM methodology places on players who have high rebounding and assist rates is gameable in this way. This isn’t to say that anyone’s intent was to artificially boost performance on a somewhat obscure metric, a practice that would essentially negate the usability of that stat. But the question of whether existing tools are up to adequately capturing the value of players in this new role is definitely raised. If it was just one or two players, that problem would be more of an oddity and argument starter than a live debate. However, more teams are operating in this heliocentric fashion now than ever before.

The Total Usage stat described above is only directly calculable from the full implementation of player tracking systems leaguewide prior to the 2013-14 season, but we can also estimate what it might have been from more standard stats:

The who’s who of players reaching this level of centrality in the past doesn’t contain any real surprises: Steve Nash, Chris Paul, LeBron, pre-Heatles Dwyane Wade, the odd AI or T-Mac super season. Iverson’s late Philly teams could actually be seen as the prototype for this sort of offense, as those teams acquired players to defend, defend some more and then go after offensive rebounds as AI was forced to be an offense unto itself.

Those were specific and sporadic squads though. This year, five teams have a mononymned star dominating the action that this degree, as in addition to the Mavs and Hawks, the Lakers (LeBron), Bucks (Giannis), Clippers (Kawhi, though one suspects this will change as he builds up floor time with Paul George), while the Rockets have two with both Harden and Westbrook clocking in at above 50 percent total usage:

In some cases, the effectiveness of this arrangement is self-explanatory. Heading into tonight’s possible Finals preview with the Clippers, Milwaukee has won 13 straight and sports the top (just) schedule-adjusted Net Rating of all time. The Lakers have gelled faster than expected and have been pretty dominant themselves. The Mavericks’ offense has been historically good so far. On the other hand, no team with a player hitting or exceeding the 50 percent Total Usage (either measured post 2013 or estimated prior) mark has won the title in that season, a seemingly paradoxical result considering the demonstrable relationship between superstar level individual production and title chances.

Advertisement

That brings me back to the Westbrook question above of what does it mean? Our existing tools for analysis of player value rest on certain assumptions about how the game is being played, but may lack utility when the game changes. Not that the 50 percent Total Usage threshold is a magical barrier, but perhaps it approximates some sort of inflection point when the game moves from one paradigm to another.

In association football (aka soccer), a ball-playing central midfielder can have outsized influence on the game, and his passing and vision can drive a team to greatness while still needing other top talent around him, but he will never have the degree of control over play as does the quarterback in American football. Perhaps we are seeing a shift from the former to the latter, which brings with it a whole new set of challenges. How do we compare a point QB to a regular old basketball player, and perhaps more importantly, will we be able to identify when it makes sense for a team to play one form of football or the other? The answer to that question will certainly be relevant to the Mavs and Hawks for years to come, but will also have tremendous import on this year’s title race.

(Photo: Austin McAfee / Icon Sportswire via Getty Images)

ncG1vNJzZmismJqutbTLnquim16YvK57kG1pcGhlbnxzfJByZmpqX2WDcMDHnmSnnadiu6OtjKGcpaGfmLKvwNGiqqZlmKTEbsDEmqSsZaKaw7C41Z5kpqeimnqivs6upZ1lpKSxosXSZqqtmaKoerW0wKdkraCVrnqmwsSrZKGZppp6o7HFqKmeZw%3D%3D