The NBA has quadrupled its 20-point scorers in nine years. How? Plus, a look at arena effects

For this week’s edition of the Analytical Lookaround, a look into two statistical oddities that I’ve been aware of.

Why are there so many 20-point scorers in the NBA right now?

Last week, while discussing some of the problems relating to statistical milestones in the crazy scoring environment we’ve seen this season, I noted 43 players were averaging at least 20 points per game this year, when that number was 11 as recently as 2012-13. A lot can change in a week, as entering Monday’s games, 44 players who had played in at least 20 games this season were doing so. Over the weekend, a reader asked what has caused this increase. It’s a great question!

There are three things converging to create such a friendly environment for big individual scoring averages. The first is pace of play. Pace as measured by possessions per 48 minutes — the definition of the “Pace” statistic found on Basketball-Reference and NBA.com — is down a tick this year at 99.5 possessions/48 after bouncing just above 100 possessions/48 for the previous two seasons according to Basketball-Reference. But those are the three fastest seasons since 1988-89. More possessions equal more shots. More shots equal more points even if efficiency isn’t going up.

However, efficiency is going up. The league-wide offensive rating was 111.8/100 entering Monday. Since the 1973-74 season, when the introduction of offensive rebounds and turnovers to the box score allowed the stat to be tracked, the prior record was 110.6/100. Last year. Effective Field Goal Percentage? 53.6 percent, a record. Free throw percentage? 77.8 percent, a record. 3-Point percentage? 36.8 percent, a record. Turnover rate? 12.6 percent, not a record, merely the second-lowest of all time to 2018-19.

The result of the league playing as fast as it has in 30 years and more efficiently than at any other point in history is predictable. The NBA’s 112.0-point scoring average is 11th all time and the highest since 1970-71. Pace stats aren’t kept that far back, but teams averaged 9.6 more field goal attempts and 10.7 more free throw attempts per game that season than they have so far this year, strongly suggesting the league played at a much higher possession rate in those days.

In 2012-13, those eleven 20 point scorers did so in a league averaging 98.1 points per game; 13.9 extra points to go around is a lot, but probably not enough to quadruple the number of 20 point scorers.

That brings us to the third leg of the stool; more concentrated offenses. I’ve talked mostly in terms of the extreme form of this concentration: heliocentrism. But a few more guys here or there making up for half of a team’s offense doesn’t get us to 44 20-point scorers. I was curious to see if there was a broader trend in terms of offenses concentrating usage more in the hands of a single player or small group of players.

There are a number of ways to look at this. This study (by Ethan Carpenter, a junior at Villanova. The kids are alright) looked at lineups in terms of broad archetypes — big threes, two-star systems, different varieties of single-star attacks and egalitarian lineups — and found that single-star lineups (featuring players with usage rates of 30 or higher with the next highest usage being no higher than 22 or 23) have been rising steadily over the last decade-plus.

I wanted to look at the panoply of line constructions more holistically. The results from a categorical approach can be heavily influenced by different decisions about endpoints.

Instead, I looked at every lineup which logged minutes starting with 2007-08, the first season NBA.com has readily available lineup data. For each lineup, I found the highest single usage rate as one measure of offensive concentration, but I wanted something a little more robust as well. To that end, I found the Gini coefficient of usage rates across each five-player unit. This analysis measures the degree of inequality within a given set of variables and returns a value between 0 (for perfect equality, every player in the lineup has the exact same usage) to 1 (perfect inequality, which isn’t actually possible across five-man units. A lineup where a single player had a usage rate of 100 — while the other four never ever shot — would return a Gini coefficient of 0.8).

The coefficient itself has little intrinsic meaning, the comparison between values can be enlightening. In this case, the first observation is that perhaps unsurprisingly, NBA lineups have a low base rate of inequality given how many usage rates cluster between around 18 and 22. That said, the average (weighted by minutes played of each lineup) amount of usage inequality has risen a decent amount over the last 14 seasons:

From 2007-08 to this season, that change represents more than a 21 percent increase in the concentration of usage within lineups. Similarly, the weighted average of the highest single usage in every lineup has risen nearly three points over the same time period.

So, more possessions, played more efficiently, with teams feeding higher volumes to their top scorers than at any time in recent history. That’s how you get to 44 20-point scorers.

On Arena Effects

A while back, The Athletic’s Wizards beat reporter Fred Katz, on his podcast, noted an odd phenomenon of Washington’s season. According to Cleaning the Glass’s “Location eFG” metric — a quick shot-quality model based merely on publicly available shot zone data — to that point in the season, the Wizards had the best opposing shot profile in the league but were also allowing the highest effective field goal percentage in the league. Though the gap has closed somewhat with the Wizards up to 25th in defensive eFG% while remaining 1st in defensive Loc eFG, this is still a chasm of a difference. Keep in mind this chasm is not quite as big as what New York is experiencing, leading the league in (non-garbage time) defensive eFG% while dead last in defensive Loc eFG% according to the CTG.

This was notable enough to look into a little more. As I wrote last week, it would be possible for a team’s results to diverge from their opponent shot profile via some combination of outlier opponent shooting and either very good or very poor rim protection. Opponents have banged home 3s at above-expected rates against the Wizards, but not to a shocking degree. Meanwhile, prior to his injury, Thomas Bryant might have been the worst rim-protecting center in the league, with Moe Wagner not far behind, so rim attempts were very profitable indeed for Washington opponents. But these factors didn’t seem like enough to account for the full discrepancy, so I remained curious.

I became even more curious when Ben Falk (creator of CTG) asked me if I thought there was something weird going on with Washington’s data, as there were some very odd home/road splits in terms of their shot location data from play-by-play. After Sunday’s game against Boston, when Washington plays at home 43.7 percent of their shots in the paint have been recorded as coming inside the restricted area. On the road, 68.1 percent. Similarly, Wizards’ opponents have been credited with taking only 33.8 percent of their paint shots in the restricted area in games played in Washington, but 64.4 percent on games played anywhere else.

As this seemed to affect both teams equally, this looked an awful lot like what those in hockey analytics might describe as “rink effects,” whereby either because of differing sightlines, vantage points or interpretation, the official scorekeeper in one building might record a shot in a slightly different location than another. Certain scorekeepers having certain idiosyncrasies isn’t a new thing. Earlier this season, Mike Beouy illustrated how the time between a missed shot and the recording of a rebound differs by arena:

The weird venue bias in rebound timing persists.

Charts below measure average time elapsed between when a miss shot is recorded and when the rebound is recorded. pic.twitter.com/hgVDnCFyQn

— Mike Beuoy (@inpredict) January 23, 2021

Much like shot location data, this has no real effect on gameplay itself but can wreak havoc with trying to determine such things as transition propensity and scoring. Most play-by-play derived transition metrics are based on the time elapsed between change of possession being recorded and the ensuing scoring action. So in an arena that recorded rebounds particularly quickly, there would appear to be somewhat fewer plays occurring in “seven seconds or less.” Perhaps coincidentally, in games played at home, Houston ranks 29th in percentage of plays in transition following a live ball rebound, but is 18th in road games.

Note there is not necessarily anything sinister going on here. Given that both this effect and the shot location differences above are affecting both teams somewhat equally, this is not a “hometown assist” situation where a friendly scorekeeper is awarding phantom stats. But it does have to be reckoned with when conducting game analysis.

In any event, I also wondered if this was akin to some of the Bubble Effects we saw during the restart and 2020 playoffs or if it was an effect that has persisted over a number of years.

First of all, I wondered if there had been some change in the underlying interpretation of when a shot is taken in the restricted area or from the floated zone of the “Paint Non-RA” as it is designated on NBA.com. Here is the leaguewide breakdown in shots from the restricted area and non-RA in terms of accuracy from each distance, the proportion of shots credited in the restricted area as well as the overall league paint field goal percentage:

Over the past decade and a half, even as there has been a shift of paint shots from the restricted area to the rest of the paint, accuracy in the paint as well as both shot zones with the key have ticked up. The reasons and mechanisms for these shifts are far too complex and interconnected to delve into in detail here, but I suspect at least some of what is happening is that the increased space and reduction in post play in today’s game has meant more paint shots are taken moving towards the rim. With that movement, there may be a little more inconsistency or confusion as to whether a shot was taken in the restricted area or just outside. Meanwhile, because of the additional space, there are fewer bodies in the way, leading to a slight uptick in finishing. At least that’s part of my theory.

So, against that general backdrop, is there anything we can see that’s counter-trend or just different in any given arena? In fact, we can. Much as there are some arenas that record rebounds “faster” or “slower,” some cities are “stingier” with credit for “restricted area” shots. And then there’s Washington this season:

Each dot represents a single season since 2004-05, with the height of the dot being the difference between the proportion of paint shots recorded as in the restricted area in a team’s home games — for and against — as compared to their road games. Over time, there have clearly been some arenas where the scorekeeper has erred more on the side of shots being “inside” the charge circle (Charlotte, Chicago, Indiana, Milwaukee and Portland stand out) while others swing the other way, notably Golden State and Sacramento but also Houston and Miami. And then there is Washington, where the discrepancy between home and road shot locations I described above breaks my whole graph.

Note that this is all from play-by-play and official NBA shot chart data, not player tracking data. That latter source has its own foibles in accounting for precise shot locations on forays to the rim, and there is not necessarily a right or wrong answer to be had anyway. But this is a good reminder that the data we use to construct our statistical narratives is often subject to almost invisible bias in how it is collected and maintained.

Also, the Wizards are probably allowing better shots than their “defensive shot profile” might otherwise suggest.

(Top Photo of Bradley Beal: Maddie Meyer/Getty Images)