Dec 2, 2014; Auburn Hills, MI, USA; Los Angeles Lakers guard Kobe Bryant (24) during the third quarter against the Detroit Pistons at The Palace of Auburn Hills. Mandatory Credit: Tim Fuller-USA TODAY Sports
Recently, I released my player consistency metric, a measurement of how likely players are to perform to our regular expectations of them from game to game. The results were fairly straightforward: there was a measure of each players’ “consistency” in each category of points, rebounds, assists, and efficiency, as well as an aggregated “variance,” a component part of consistency[1. The methodology behind the consistency metric can be found in the link above. Player consistency will be regularly updated and can be found in the Our-Stats page on the site at any time.]. That metric was thanks to, almost entirely, a methodology given to me by co-writer at Nylon Calculus Krishna Narsu, without whom consistency wouldn’t work very well.
It occurred to me, though, before publishing my first real update of the data, that two things were missing: first, an aggregated, general “consistency,” like the aggregated “variance,” and an aggregation method that made any real-world, applicable sense beyond “well, I through the numbers together and it looks like this jumbled mess,” which is basically what I did with the variance[2. There was a methodology to my aggregation of the variance, but it was mostly built around fantasy sport aggregations, which has no meaningful application to what’s happening on the court. In short, I put the numbers together in a way that reflects the player’s consistency regarding their contribution to the box score, but not in a way that reflects their contribution to team success. That has, thankfully, been addressed.].
I did, though, for the first major update of the consistency data, address this, allowing me to present the data along with a more general “consistency” column that incorporates a player’s scoring, efficiency, rebounding, and passing consistency into one, meaningful measure! I adjusted the variance aggregation, too, to operate roughly the same way.
The newly aggregated consistency and variance metrics are averages of the points, minutes, 3PM, FG%, rebounds, and assists consistencies and variances, but weighted in a manner proportionate to the SPM used by Nylon Calculus co-writer Kevin Ferrigan. The idea of an SPM is to use box-score metrics (like the ones upon which consistency is based) to approximate “on-court-impact,” as measured by APM’s like RAPM or RPM. A great explanation of what APMs are, and why they measure on-court impact, can be found here.
In other words, I have averaged out the various consistencies in a way that reflects each one’s importance to a a player’s contribution on the court.
The result of that averaging is as follows, and it will, as before, be updated on the Our-Stats tab on the homepage:
While you’re here, I thought I’d take some time to share some fun tidbits with you about the early returns of consistency, now that we’re nearing a sufficient sample size for meaningful discussion.
The top 10 most consistent players in the NBA right now, among players who have played significant minutes and who have a low variance (i.e. we don’t already expect them to have pretty wildly varied games from game to game), are:
- Pero Antic
- Marvin Williams
- Wayne Ellington
- Kirk Hinrich
- Matt Barnes
- Pablo Prigioni
- Aaron Brooks
- Jason Terry
- Devin Harris
- Eric Gordon
What I find particularly funny about this list — and it is funny — is that most of these players (Brooks and Harris excepted) are bad or, at best, mediocre at the NBA level. The average FG% — which the most strongly weighted “consistency” by far — of this group is like 42%. But consistency isn’t about “good” or “bad,” it’s about regularity, and these players are the most regularly bad or regularly mediocre of any player in the league.
There’s probably an interesting question to be asked about why bad players are more prone to higher consistency, but that’s another question for another time.
On the other end of the spectrum the least consistent players, given significant minutes and a high-ish variance, are:
- Timofey Mozgov
- Kenneth Faried
- Henry Sims
- Dennis Schroder
- Al Horford
- Luis Scola
- Giannis Antetokounmpo
- Jarrett Jack
- Omri Casspi
- Blake Griffin
Man, the Nuggets are having a rough year. Of these players, Mozgov, Faried, Sims, and Schroder all had a negative consistency, which is extremely hard to do over significant minutes. Blake’s consistency was hurt badly by a disastrous three-point consistency from having two games with decent three-point performances and then no other threes at all, and his consistency leaps to the top of the league without the inclusion of his odd three-point adventure. If “a significant sample of threes” was required, Doug McDermott would replace Blake as 10th on this list.
Among top scorers, the top three most consistent at scoring in every game are:
- Dwyane Wade
- Carmelo Anthony
- Anthony Davis
While the top three in consistent efficiency among top scorers are, funnily enough:
- Kobe Bryant
- Kyrie Irving
- Dwayne Wade
Once again, the scorers who are the most inefficient are the most insistent on staying inefficient in every single game.
There’s lots to look at and play around with, though, so enjoy, and feel free to send me anything fun that you find on twitter at @HalBrownNBA!