Unsustainable Starts: Rockets, Raptors Candidates for NBA Regression

facebooktwitterreddit

Nov 22, 2014; Houston, TX, USA; Dallas Mavericks forward Dirk Nowitzki (41) shoot over Houston Rockets guard James Harden (13) during the second half at the Toyota Center. The Rockets defeated the Mavericks 95-92. Mandatory Credit: Jerome Miron-USA TODAY Sports

Almost 20% of the way into the NBA season, it’s probably ok to stop saying teams have “started well” but rather just describe them as being good. Certainly, teams will ebb and flow over the year with roster moves and especially injuries altering performances greatly. But teams are starting to show what they are and what we can expect from them over the rest of the year. That isn’t to say we can’t make some educated guesses about the how teams might regress towards the mean in specific areas however.

With the finely detailed shot data now available on NBA.com, a much better understanding is available than in years past about what might broadly be described as a team’s shot quality. Earlier this week, I noted there is more information than ever on what is and isn’t a “good” shot. This chart is illustrative:

To some degree, a reconfirmation of things we already knew: midrange bad, open shots good. Going off of this broad data, it’s possible to examine which teams are prospering because they are “getting better” shots and which are simply better at making the shots they are getting. The quick and dirty version is represented here as Expected Effective Field Goal% (XeFG), and is simply what the team would be expected to shoot given the mix of shots in across the three major shot zones (close shots, mid range and three pointers) across varying ranges of defensive pressure[1. Because so few three pointers have been taken with “tight” contests, meaning the defender within 2 feet of the shooter, “tight” and “moderate” three point defense were combined into one column for simplicity to measure all three pointers where the defender is within 4 feet of the shooter.]. By comparing this to the team’s actual shooting,[2. A few games and stray shots appear to be missing from the underlying data set, which accounts for some minor differences between this data and the overall season data found either on B-ref or NBA.COM] the difference between expected and actual shooting can be discerned:[3. To a large degree, this method is a natural extension of

Ian’s Expected Points Per Shot metric

taking into account defensive pressure.]

While it’s probably too early to draw general conclusions from this data, some possible trends are discernable. For example, Golden State is outperforming their XeFG% largely on the strength of three point shooting. The Warriors are hitting wide open threes at just above NBA average rates (39.7% vs. 39.3% for league as a whole), but they are also hitting more contested 3s at an extremely high rate, shooting 39.5% on moderately open 3s (NBA average 35%) and 38.4$ on contested 3s (30.8% leaguewide). To some degree that is likely to come down a bit – only one team in the league hit above 34% on contested three last year, and that was the Spurs at 36.5% . But with Klay Thompson and especially Steph Curry around, of course Golden State is going to hit contested 3s at a high rate, they did so last year as well, hitting defended threes around at about 2.5% higher clip than league average. On the other hand, Toronto’s explosive start to the season is probably not sustainable. Thru Thanksgiving, the Raps were hitting an absurd 50% of their contested three attempts. The Raptors shot 28.7% on such shots last season, and while Lou Williams is a maker of tough shots, not that much of one, because nobody is.

A similar analysis is possible on the defensive side of the ball to see which teams are have been good at forcing opposing offenses into bad shots and which teams have been more or less effective at contesting those shots which have been taken:

Much like as with the look at team offense above, determining what to make of some of these numbers is tough, but one thing pops out right away looking into the more detailed data: Houston’s defense has been a mirage to some degree. Tops of the league in terms of the biggest gap between their xEFG% and actual eFG% allowed, the Rockets performance is driven by three point defense. Opponents are shooting 27.5% from three against Houston.

While one interpretation of this might be the MoreyBall-inspired Rockets know the value of the three so don’t allow as many open threes, the detailed data doesn’t really bear this theory out. The Rockets are slightly better than average in terms of being in position to contest 3 point shooters but only just. The Rockets allow opponents to shoot right around league average proportion of shots from deep, with more or less proportional defensive attention.[4. A slightly higher percentage of Houston’s opponent’s 3s occur with a defender between 4 and 6 feet away, with a commensurate drop in those coming with a defender more than 6 feet away.]

What seems to be happening is guys are just missing shots. Houston’s opponents are shooting under 32% on wide open threes, under 26% on moderately open threes and a laughable 19.3% on contested threes. In fact, the three teams most outperforming their XeFG% defensively are those whose opponents’ are collectively shooting the worst on wide open threes, with Portland and Oklahoma City also “holding” shooters to under 32% when wide open beyond the arc. While it’s possible (especially in the case of an extremely long and rangy defensive team like the Thunder), no team held opponents below 34.7% on wide open threes last year and neither Houston nor OKC were particularly good on this front a season ago.[5. Portland’s 36.3% was good enough for 4th best.]

Much like the Raptors contested shooting, the likely coming regression doesn’t mean Houston’s opponents are going to suddenly morph into a terrifying mix of Kyle Korver and Steph Curry, but opponents merely shooting as expected for various levels of defense would be enough to take Houston from their current 3rd best eFG% allowed to around 12th. Of course some of the deviation from the norm could be legitimate skill differences, and only the rest of the season will demonstrate conclusively either way, but for now, I expect both Houston and Toronto to drop off a little from their level of play to this point in the season.

This is just a quick and dirty look at some of the data, if you want to look for yourself, I’ve uploaded team’s offensive and defensive efficiencies for various shots and degrees of defense to google docs here.