Data Mining: How Wilson Chandler is (and isn’t) regressing to the mean

As the Nuggets got off to a rocky start to begin the season, it was hoped that help would be on the way in the form of Wilson Chandler when he returned to play after missing their first six games due to a hamstring injury. And while he has made helpful contributions to an extent, he has clearly fallen short of making the impact many Nuggets fans were hoping for.

For starters, his impact on the court versus off has been essentially neutral. Consider, for example, the discrepancy between Chandler and Mozgov in this regard:

With Chandler, the difference between him being in the game or not has been negligible. The Nuggets have netted just 0.8 more points per 100 possessions and have shot only a 0.6 higher TS% when he’s been on the court. By contrast, with Mozgov in the game Denver has netted a significantly higher 6.4 points per 100 possessions, and shot a 5.0 percent better TS%. To put this into perspective, Denver currently has a pace of 98.5, which is close enough to 100 for us to say that Mozgov is netting just over plus-six points per game, while Chandler is at just under plus-one.

Mozgov is making the kind of big positive impact which had been expected from Chandler, but has so far not really materialized. And the fact that Wilson has yet to break through his early season struggles (as, say, Nate Robinson did) has been one of the more disappointing aspects of this Nuggets season. Perhaps his first few games back from injury might be dismissed due to recovery time, but now that he is a full 15 games* into the season and appears to have been at full health for most of that stretch, it’s becoming apparent that the version of Chandler the Nuggets are getting this season is not quite up to par with last year’s model.

 

Regression to the mean, shot selection or both?

One of the more glaring dropoffs in Chandler’s game this season has been in his 3-point shooting percentage, leading many (including myself) to speculate that last season may have been a fluke and that he is now, as could be expected, regressing to the mean. And while there appears to be some truth to that, there are other factors (as we will see below) which may be playing a greater role in Wilson’s reduced efficiency.

But before diving into the deeper complexities, let’s first address the main question at hand: Is Wilson Chandler’s 3-point shooting regressing to the mean?

In a word, yes. And it doesn’t take much more than a look at this comparison of his shooting percentages to see it quite clearly. The dark blue bar on the left represents the average from his career prior to the 2012-13 season, the light blue bar is last season, and the yellow bar is 2013-14:

You can see that both his 2-point and overall field goal percentages are within the ballpark of his career averages, but that his 3-point percentage spikes to over .400 in 2012-13 and then plummets back to the .350 neighborhood of his career average this season. So there’s the regression, and unfortunately that seems to make it unlikely that Wilson can find a way to re-create his better 3-point percentage moving forward.

There is another interesting point to be made here, which is that despite the slippage in his 3-point shooting, Chandler’s eFG% is just as high as last season’s, his TS% has dipped only slightly, and both are above his career averages. We will return to this apparent incongruity further below.

Before that, however, let’s take a look at a comparison between this season and last of the types of plays Chandler has been making, as this will provide some context and detail for the regression:

As you can see, there has been a dramatic difference in his play type profile this season. His spot-ups (largely 3-point attempts) are way up, while his plays as the pick-and-roll roll man (generally drives to the basket) are way down. And while it’s true that, all things being equal, both 3-pointers and shots at the rim are among the most efficient shots in basketball, all things are not, in fact, equal:

Chandler has been highly efficient in pick-and-rolls this season, scoring 1.6 PPP as the roll man. And while we can expect this, like his 3-point shooting, to see at least some regression as the sample size grows through the season, when considering the possibility that his 3-point shooting has already regressed further, the result so far is that he has been taking his less efficient (ie. lower PPP) shot (3-pointers) more frequently and his more efficient shot (drives as the roll man) less frequently. And so, on the balance, shot selection may be playing a role just as important as, if not greater than regression.

 

How usage factors in

Let’s turn to Chandler’s actual production on the court, in this case comparing his per 36 minutes statistics along the same career/2012-13/2013-14 divisions:

A quick glance reveals that his production is down from last season nearly across the board, except for one particular area: made and attempted 3-pointers. Despite a drop from 15.2 to 11.4 overall field goal attempts per 36 minutes, Chandler’s 3-point attempts have jumped up from 4.2 to 5.6 this season. And consequently, his 2-point attempts have plummeted from 11 to 5.8. If his 3-point percentage is regressing to the mean, his attempts are doing quite the opposite.

This provides some explanation for the question posed above as to how he could maintain his eFG% despite the drop in his 3-point percentage: It is being offset by an increased volume of higher value shots. Last season, about 49 percent of Chandler’s shots were taken at the rim. That has dropped to about 29 percent this season, and one would think that would lead to a big hit to his efficiency. Meanwhile, however, the percentage of his shots taken from the corner has increased from about 11 to about 20 percent of his total shooting. This appears to be enough to offset the drop in his 3-point percentage.

However, there are other consequences of staying out of the paint and sticking to the perimeter. His rebounding and free throw attempts are also down, and the latter – along with the drop in his free throw percentage from .793 to .714 – is why his TS% (unlike his eFG%) has slipped.

Let’s take a look at one last chart which will help to bring this all together:

Here we can see in striking relief the trends described above. Chandler’s PER, free throw rate, rebound rate and usage have all dropped off, while his 3-point rate has skyrocketed.

But although the surge in his 3-point rate may be the most eye-catching, I’d like to end by proposing that, perhaps, the drop in usage rate is the most important stat of all we have looked at. Brian Shaw has said repeatedly that he wants to find ways to get Chandler more shots, and get him more involved in general, but whether it’s due to Shaw’s coaching, Chandler’s play, or both, they have not had much success making that happen in reality, as reflected by Wilson’s career-low usage rate of 16.8.

Chandler has looked out of place and confused about his role this entire season. Shaw hasn’t really been running many set plays for him such as the “pick and fade” George Karl went to so often last season (see Mike Prada’s excellent breakdown of that here). Chandler’s greatest offensive attribute might be his versatility, but that has largely been squelched in favor of just bombing from the arc all the time, which has produced neutral, but not fruitful results.

I’m not optimistic that Chandler can reverse the regression to poorer 3-point shooting, but steps can certainly be taken to improve his overall efficiency. If Shaw wants him to be more involved, he needs to put his money where his mouth is and draw up some pick-and-rolls and pick-and-pops for him. Chandler plays better when he’s more aggressive, and getting him to the rim more often should not only help his confidence, but also his free throw and rebounding rates.

And who knows. If he’s more selective about his 3-pointers, and keeps defenses honest by driving to the rim more, maybe, just maybe he can start working on fighting his way back upward against that regression and closer to .400 again.

 

You can follow Joel on Twitter here: @denbutsu

 

*The statistics in this post were compiled prior to the Dec. 13 game versus the Utah Jazz, and are courtesy of Basketball-Reference.com, MySynergySports.com, nbawowy.com and NBA.com/stats.

 

  • David Acker

    While I’m not ready to give Chandler a pass he clearly is not 100%. During a lot of his rest periods he gets on the bike to keep his groin warmed up. While listening to the opposition announcers ay that his groin still gets tired. Wilson is a nice player to have coming of off the bench. When asked to start he normally gets a lock down defensive role as well.

  • Charliemyboy

    My immediate response is that his results are Shaw’s system and a little hesitancy to rim attacks due to his injuries. He is told to shoot the 3 rather than drive like he was told to do last year. The entire team is that way. We don’t command the lane anymore (we don’t command anything). Again, Karl used the talent he had; I have yet to see that currently with maybe only Moz and Ham taking advantage. Same with Chandler’s rebounds; how can he get more rebounds shooting 3′s? It’s very complicated; up in one area, down in another. But the genius is in understanding what works more of the time. I am surprised at how mediocre teams are getting better and so are better teams. We better get with it, or will be left behind this year. No one on this team has the talent or ego to be mean enough to demand wins.

  • Kyle Wurtz

    First of all, really great stuff. I love digging through the numbers like this, as it often confirms suspicions we have while watching the games. And I think that you hit the nail on the head that Chandler isn’t playing at the level we really need him to play.

    One thing to consider, though…

    When looking at the on/off court summaries, comparing the values for a starter and a bench player, especially for a deep team like the Nuggets that relies on its depth to compensate for a less talented (relative to its opponent) starting lineup, isn’t really appropriate IMO. You don’t have to look any further than Ty and Dre Miller to see this. The Net Rating (using this stat for simplicity: http://stats.nba.com/teamOnOffSummary.html?TeamID=1610612743) for the team while Ty is on the court is -1.9. The only player worse is, surprise, JJ Hickson at -5.4. The Nuggets’ Net Rating with Miller on the court is a whopping 7.3.

    The reverse statistics (Nuggets’ Net Rating while the players are off the court) are 11.9 for Ty (Nuggets are 11.9 points per 100 possessions better with Ty off the court) and -0.5 for Miller (Nuggets are actually worse with Miller off the court). Does this mean that Ty is a significantly worse player than Miller? Obviously not. Miller spends most of his time playing second-unit time when the competition is less than par. Miller’s not going to spend much time guarding Chris Paul. There’s a similar situation with the Chandler/Mozgov situation. Chandler spends a lot of time guarding Dirk and KD…Mozgov doesn’t. It’s hard to compare those two.

    So, yeah, Chandler’s stats don’t jump out as being awesome. But his stats are probably skewed because he’s playing against above average opponents, and he’s probably having a much greater positive impact to the team than we’re giving him credit for when you consider the team’s other options (cough, Anthony Randolph)…

    • Evan Woodruff

      Very very very very good point.

    • http://www.roundballminingcompany.com/ JoelRMC

      Yes, very good point indeed, and I concede it completely. And thanks for bringing it to my attention. I had not thoroughly thought through the implications of comparing a starter vs bench player (my head was mostly wrapped up in sorting through the stats that followed), but you’re absolutely right. In fact, after reading your comment I went back and did up the same numbers (net differential of points per 100 possessions and TS%) for all the players currently in the rotation, and the results are in the attached image.

      Of course, as you point out, it would be ridiculous to say that because Ty’s at the bottom of the list that he’s the worst player on the team. And an easy statistical counterpoint would be that he leads the team in PER and win shares.

      At any rate, though, I think the salient point that I was trying to make — that Chandler was performing more poorly than expected — is essentially true. But if the fact that his net PPP & TS% are the best among the starters *despite* more limited offensive production tells us anything, it’s probably that his work on the defensive end is making up for it to a certain extent, and he probably does deserve more credit for that (including from me).

  • Dan Mikanmaster

    Good post. That pick and fade you mention was especially useful when WC was at the 4. This year with our plethora of PF’s he hasn’t had much time there.

    One note – text-heavy images are better saved as png’s than jpg’s. The text doesn’t get distorted like the jpeg compression does.

    • http://www.roundballminingcompany.com/ JoelRMC

      Thanks for the suggestion. I tried it with the chart I posted in my comment above. Does it look better to you? I really have no technical knowhow about video and image editing. All I know I’ve figured out on my own, so there is a LOT I don’t know, and I’m sure some of what I think I’ve learned is probably wrong. Also, my vision isn’t so great, especially when it comes to discerning fine detail, so there may just be times when I don’t notice it could look better. Anyhow, thanks for the tip.

  • The Truth

    Nice how numbers don’t lie and don’t tell the whole story at the same time. If he’d hit that 3 late against Utah, the game, the rapid reaction, my post to that thread and this post would all likely be different or at least received different reactions.

    So many things are part of winning and being a good player that are not often discussed. Stemming the opposition’s momentum and making timely plays are two that are hard to get data on. Stuff like this could use some context like the earlier reply referencing Ty’s numbers. Good to see Mozzy eat up most of his bench competition in any event.

    Gallo and Wilson are the same player to me, both need to learn how to go to the whole and finish like the near 7 footers they are (I wish they would gather and go off of two feet like Melo does alot). Both are better than average 3′s with the ability to play 4 and stretch a D. The biggest difference between the two is their fan base, and who gets the benefit of the doubt among most coaches, pundits and fans.

  • JED89

    Wilson’s tenure in NY has been unfortunate and I really think he will show us by the end of the year just how dynamic of a player he can be. He was an 16 and 6 player in NY and I think he can match that here if not due better. He was clearly not himself when coming back but look at his last 4 games, 15.75 points, 6 rebounds 2 assists. These are not all star numbers but with an allstar TY, a healthy Gallo and maybe another piece or two so Wilson is coming off the bench and we are a dangerous team. Wilson is also an intense and versatile defender, his numbers and PER aren’t gonna reflect that. He and Gallo will be a great combo next to Ty. Hopefully the Knicks continue to suck and we can get one of the 7 studs in the draft next year.