Nylon Calculus: Are NBA superstars better than ever?

Feb 1, 2017; Oklahoma City, OK, USA; Oklahoma City Thunder guard Russell Westbrook (0) reacts after a play against the Chicago Bulls during the second quarter at Chesapeake Energy Arena. Mandatory Credit: Mark D. Smith-USA TODAY Sports
Feb 1, 2017; Oklahoma City, OK, USA; Oklahoma City Thunder guard Russell Westbrook (0) reacts after a play against the Chicago Bulls during the second quarter at Chesapeake Energy Arena. Mandatory Credit: Mark D. Smith-USA TODAY Sports /

This season the NBA has seen an explosion of stat lines unlike anything we’ve seen before. Players are casually dropping 40 point triple-doubles — There were 19 such instances of 40 point triple-doubles from 1983-84 to 2015-16. In this year alone, there have been 8 games, all by Russell Westbrook and James Harden. From the absurd stat lines of Westbrook, Harden, Giannis Antetokounmpo and more it seems like new statistical ground is being broken everywhere we look. But is this true? Are the stars at the top of the league better than ever?

To determine this, I looked at both the average and weighted average (by relative rank, first = X *1, second = X*0.9). etc.) of the top ten players in each season for various all-in-one metrics (PER, WS/48, BPM, and HBox).


weighted-avg-bpm-of-top-10-players /

One potential issue with looking at BPM is that it was trained on data from 2001-2014. So modern superstars may get a slight edge in that the coefficients may be different if the data was trained on the 1970s-80s (where for example, offensive rebounding and free throw rate were more important due to the league average being higher.).

However, I don’t really think this is an issue. BPM adjusts for the league average 3-point rate, which is where we would see the biggest jump for modern players. But it is worth pointing out. We can look at PER, where the league average is set to 15 every year:

weighted-avg-per-of-top-10-players /

Like with BPM, we see this year has the highest average PER. In fact, there are currently nine players with a PER over 27. The previous high in a season was four (last year and in 2005-06). The issues with PER are well documented (and we’ve even tried to fix it) but it fits the same pattern as the other metrics here.

And here’s Win Shares per 48:

weighted-avg-ws48-of-top-10-players /

Finally, we can look at HBox. You can find an explanation for HBox here and here. One advantage of HBox is that the data is trained on all seasons since 1974, so there isn’t going to be any bias towards modern players. Additionally, the metric adjusts for the league average true shooting percentage so modern players won’t get a boost for playing in a more shooting friendly environment. Although with the exception of the last two years, the league average TS% hasn’t really changed much over time.

weighted-avg-hbox-of-top-10-players /

As we have seen throughout all of these metrics, the best players in the league appear to be better than ever. There has been a steady increase since the 2010-11 season with the league reaching its apex in this current season. So we appear to have our best set of superstars in league history.

What is the reason for this? I took a look at the Shot Usage and TS%+[7.  TS%+ = true shooting percentage – league average true shooting percentage.] of the top 10 players in HBox for each season to see if there has been an increase over time.

First, TS%+ for the top 10 players by HBox each season:

avg-ts-pct-of-top-10-players-in-hbox /

We don’t really see any change in TS%+ for the last few years. So this indicates that the modern superstars aren’t really getting any more efficient relative to league average. Interestingly, we do see that there was a spike in the early 1990s.

Here’s Shot Usage:

avg-shot-usage-of-top-10-players-in-hbox /

Here is where we see the change. The top players in the league are taking more shots compared to any other time in the league. Given the fact that their TS%+ stayed relatively constant over time, this indicates that the increase in these modern superstars’ all-in-one metrics may be a result of their increased usage.

Next: A new twist on measuring team 3-point defense

Finally, while we do see that the top players in the league are better than ever, I think it is worth mentioning that many of these metrics, which are all derived from box score components, are more offensive oriented because of the limited defensive box score statistics that are available. And as both Zach Lowe and Positive Residual have pointed out — offense has been up this season. So many of these superstars are stuffing the box score with offensive stats in an offensive environment. And it’s possible that we aren’t capturing the defensive value appropriately (this goes for players throughout history).