Earlier this fall, we introduced non-negative matrix factorization to help us analyze field goal trends within the NBA and applied it to the 2017-18 NBA season. There we compared the Boston Celtics and Houston Rockets to show the immense discipline the Rockets displayed while playing “Moreyball.” That is, a high propensity to take three point and/or rim attempts in an effort to maximize effective field goal percentage. You can read the results here. For last year, we saw yet another leap in three point shooting, however, we barely saw the decline in mid-range shooting. It was in fact that most teams were almost uniformly borrowing from rim, paint, and mid-range attempts to bolster their three point frequency. The question is, with the Rockets’ success has the league actually made the push to play “catch up.”
Changes in Shooting Trends
After 278 games played through November 24th, we have had the chance to witness every team play at least 16 games; as the Detroit Pistons have been lagging behind most teams that have played 18-19 games by this point. This gives us the ability to gte above the noise floor and start comparing trends between this season and last season. To do this, we simply apply the non-negative matrix factorization process to the entire season and compare the spatial basis functions. And here’s what we found out…
Spatial Basis 1: Right-Hand Rim Attempts over Right-Hand Threes
We find that the primary action so far this year has been rim attacks; which is a major change from last season. Last year, the league was dominated by three and rim players. These players ranged from your driving players such as James Harden of Houston and Steph Curry of Golden State to your catch-and-shoot and slip players such as Klay Thompson of Golden State and, yes… Al Horford of Boston.
Using the Milwaukee Bucks’ court as the backdrop for this season, we see the huge discrepancy between this trend. Notice the right-handedness of last year. This will become an important factor very soon…
Spatial Basis 2: Right-Hand Action over Mid-Range
Here, we see the mid-range start to fall off in favor of a right-hand dominant game. We see the three-and-rim attack as in the prior season; even with a flair of the right-hand dominance from the previous season (see the Rockets court above). However, this season is more pronounced; which may be due to only 278 games played as opposed to 1230 games.
Note that the mid-range game is not in the top two spatial basis; which indicates a current slight decline in preference for the mid-range. In particular, the post spot-up type shots and the baseline drive-and-pull-up.
Spatial Basis 3: Threes over Dunks
By the third spatial basis, we begin to see three point attempts jump back into the mix. We also see that spatial basis three of last year (rim attempts) are currently this year’s favorite style of field goal attempt.
Spatial Basis 4: Dunks over the Right-Hand Post Game
Here, we are seeing effectively dunks as the fourth most prominent field goal attempt, whereas last year, the right-handed “Kosta Koufos” style of play was the fourth most dominant style of field goal attempts.
As we have seen this fourth spatial basis drop from last year, this is again indication of the growing “no-man’s land” that is the space between the restricted area and the three point line.
Spatial Basis 5: Threes are Back!
We see near identical favoritism to the three point line in spatial basis five, indicating that this type of field goal attempt is still a favorite among NBA teams. This makes sense given the three-point revolution over the past few years.
In the current trend, we do see some left-hand style of play and a slight decrease at the top of the arc. The top of the arc fluctuations are most likely due to noise, while the left-handed attack at the rim is most likely compensation for the right-handed dominance from above combined with the tendency for players to attack the rim and shoot the three. That is, this is separating old Spatial Basis 1 (Houston plot) and Spatial Basis 5 (Houston plot) into two re-factored types with Spatial Basis 1 (Milwaukee plot) and Spatial Basis 5 (Milwaukee plot)
Spatial Basis 6: Clean-Up of 1 and 5 versus Corner Threes
This is an odd result that pops up. Here, we see the well-known “Corner Three,” which is considered the “easiest three point shot in the game.” It has yet to be seen over the course of this season as a preferred style of shot attempt. Instead for this season, we see an artifact-laden plot where the mid-range game is ever-so-slightly introduced with free throw and elbow jumpers. We see some faint three point trend, with an ever-so-faint corner three attempt.
This indicates the artifacts are finding player with a high propensity to attack the rim while taking shots from the elblows, free throw line, and the occasional corner three. These style of players are your Kyrie Irving’s of Boston, but also players such as Mo Bamba from Orlando.
Spatial Basis 7: Corner Three’s over the Mid-Range
Here, we once again see the mid-range fall on this list. And finally, we obtain the corner three assault from teams for this season. This shows that the corner three is still indeed a prevalent part of NBA team’s systems.
Spatial Basis 8: Left-Handed Post Action, Just Closer…
Here we see a match-up of left-handed post play. The key difference here is that the action is tighter around the rim, indicating that players are more willing to either take the extra step, or position themselves, to be within five feet of the rim as opposed to be between 5 and 15 feet from the rim.
As we have seen over the years, field goal percentages drop significantly after players wander beyond six feet from the rim; in particular at ten feet do percentages really drop off. This plot indicates that both players and coaches are understanding that key point and making it an established style of field goal attempt.
Spatial Basis 9: Top of the Arc, Just Tighter…
Here’s another basis-to-basis match with the top-of-the-arc three point attempts. The core difference between these two plots is that players are taking the shots closer to the top of the arc as opposed to having a 15 foot window along the arc.
Some of this may be due to noise while some of this may be due to spacing effects of the team. Regardless, the style of field goal attempt is still in vogue for this season, at the same prevalence.
Spatial Basis 10: Mid-Range finally Makes and Appearance.
Finally, the mid-range game makes an appearance in the league. It should be noted that in a ten-factor decomposition, the tenth factor contains the “remainder” of the spatial noise. This indicates that the mid-range game is effectively the least favored shot style.
That said, its prevalence identifies that it is still a key part of the game. And while the mid-range is considered a low-yielding type of field goal attempt, the fact that players take them induces a “game theory” attack on their defenders. Read that as hitting an occasional mid-range jumper forces defenses to address the mid-range game; allowing spacing to occur on offense. The mid-range game is still indeed important to spacing.
What we ultimately find out here is that the mid-range game is indeed disappearing from the game. It has been written a lot about the transformation the Milwaukee Bucks have made over this year with Mike Budenholzer at the helm. You can find some of those articles here, here, and here. However, these articles just look at the basics of “More Threes!” and “Less Mid-Range!” But what are the Bucks actually taking for shots? We can quantify this. And within a couple of those articles, a comparison is made to the Houston Rockets. We can quantify that too.
Transformation of the Bucks
Last year, under Jason Kidd and company, the Bucks favored an all-over-the-court style of game. It was called “position-less” basketball. What it really did was force the team to shoot all over the map in an undisciplined style of play. We see that with the shot-chart from last season:
We see that shots are indeed taken everywhere on this court. A strange occurrence is that right-direction play for the Bucks actually sees some consideration of not taking a “foot-over-the-line” two-point attempt. However, in the left-hand direction… it’s out the window. This is just a strange coincidence.
Let’s compare that chart to this year:
Just slightly under a quarter of the way through the season, we see the Bucks taking a few mid-range shots; but nowhere close to the volume as before. We still see a significant amount of field goal attempts lingering in the high post/paint; but that’s an artifact of players such as Khris Middleton and Eric Bledsoe attacking from the three point line. Most important to note, there are only a couple handfuls of mid-range jumpers to the wings/baseline.
To understand this in terms of style of play, we look a the coefficients of the Bucks in terms of the spatial components. Remember, we need the breakdown above to be able to understand what is really going on.
Let’s quickly recall what those ten components were…
- Threes: 1 (right-handed), 5 (spread/wing), 6 (corner) , 9 (top)
- Mid-Range: 2 (baseline), 4 (right post), 7 (pullup), 8 (right post)
- Rim: 3, 10
OK, so what did we see last year for the Bucks? We saw a significant Khris Middleton mid-range (baseline) game. This was far and away the most preferred conditional attack from the Bucks. As Middleton attempted 15.5 field goal attempts per game (second to Antetokounmpo at 18.7), 8.37 attempts came from the mid-range. That’s 54 percent of Middleton’s shots came from low-eFG% regions. With an eFG% of .524 last season, a field goal attempt from Middleton was worth approximately 1.048 points per chance; a little below the desired 1.10 rate. This placed Middleton at 43rd in the NBA, despite the low eFG% rates; which indicates that Middleton is most likely a well-above average mid-range shooter.
If we take a closer look at the Bucks’ numbers, we do indeed see a high number of zeros on the board (indicating roles of shooters), but those numbers are associated with low-minutes players. This means most zeros on the table are structural zeros and have no influence on the style of play. Instead, we focus on players who play significant amounts of time.
If we look at Eric Bledsoe, Giannis Antetokounmpo, Malcolm, Brogdon, Khris Middleton, Jabari Parker, Brandon Jennings, Jason Terry, Thon Maker, and Matthew Dellavedova; we see every single one of these players have numbers all over the board. This indicates that the Bucks are playing the older style of basketball, allowing shooters to shoot all over the map; for better/for worse.
Taking a closer look, basis functions 2 (midrange baseline) is the Bucks’ most prominent field goal attempt type. This is followed by basis function 5, the spread wing three point attempt; and spatial basis 3, the primary rim action. This shows Milwaukee as playing a very late-90’s to mid-2000’s style of play. Don’t believe us? Here’s the 2004-05 season:
Let’s compare that to this season:
Here we begin to see the separation of shot types. To help ourselves again, we focus on what each spatial component represents:
- Threes: 3 (wing), 5 (spread/wing), 7 (corner) , 9 (top)
- Mid-Range: 8 (post action), 10 (all mid-range)
- Rim: 1, 4, 6
- Right-Handed Action: 2
Immediately we see a lot of zeros for high usages players and they are all in components 8 and 10. Look above what those components mean. That’s right. The mid-range game is indeed gone. And Khris Middleton? Almost nothing from full mid-range. Hands down the biggest change of all. And his shooting has so far improved .045 for eFG%. That’s an extra .09 points per chance for Middleton. Remember that desired 1.10 points per chance that only about 30 players in the league can attain? Middleton is at 1.13 points per chance when it comes to field goal attempts.
We also see other trends pop out: Giannis is not a corner three shooter. Donte DiVincenzo and Tony Snell are effectively the only wing-three shooters on the team and they play a limited role. In fact, if the Bucks are taking threes, they are spread about the high wing from Brook Lopez, Khris Middleton, and Eric Bledsoe. And if Giannis gets into the half-court offense, be prepared for a right-handed game. That’s the primary action for the Bucks; and no one else on the team really does it.
As we compare the current Bucks to the previous year Bucks, we see that the roster has very little turnover. Effectively the addition of Brook Lopez over Jabari Parker and some deeper rotation shuffles is all that exists. Therefore, it is safe to imply that the style of play change is due to the change of coaching.
Compare to the Rockets
Recall that last season the Rockets were by and far the most disciplined team in the league. We saw this with large blocks of zeros in the spatial components.
However, the Rockets roster turned over quite heavily from last season as we saw the departures of Trevor Ariza to Phoenix and Luc Mbah a Moute back to Los Angeles. We saw the introduction of Carmelo Anthony for a short period of time and have welcomed new players such as Michael Carter-Williams and Gary Clark into the mix. The question was whether the players could adapt to new roles or fit into the system; or if the system would change to adapt to the players. So far this season, after 17 games, the Rockets have a distribution as follows:
We see some similar trends as the previous year, but we also see a lack of discipline compared to last year. While the midrange game has vanished as a whole from across the league, we see Houston pick up the post / high paint mid-range game (spatial component 8) and it is being adopted by Eric Gordon and Michael Carter-Williams. The latter can be due to learning the system; the former is definitely not.
We also see the impact of Carmelo Anthony with the high mid-range game, rivaling that of Chris Paul; who was effectively the only Rocket (other than Harden) given the green light in this area last season. Whereas Paul was effective from the mid-range (much like Khris Middleton), Carmelo was not.
For top three’s we still see Harden, Paul, and PJ Tucker maintain their roles. However, with the departure of Trevor Ariza, we see that fourth player disappear from the ranks. This is a roster construct as neither Anthony nor Carter-Williams fit that style of shooting capability. Ariza’s shooting role has since fallen onto Gerald Green. More curious, however, we see that Eric Gordon disappear from the ninth component.
Breaking down Gordon’s role in the offense, we see that he has yet to attempt a field goal between 16 feet and the three point line. Which is the least amount he has taken ever, but we also see that Gordon is down seven percent when it comes to three point frequency. And this may be a direct result of his unexpected dip in three point percentage; .243!!!
Last year, Gordon was primarily a spread/wing three point shooter and then a corner spot-up shooter; moreso on the left side than right as he has no spatial component for right-dominant threes. Compare that to this year, Gordon’s threes have been exclusively a wing and spread/wing shooter as he has no spatial component for top threes nor corner threes. This indicates a change in the style of play for the Rockets between last year and this year; a given with quite the overhaul of the roster.
But as D’Antoni mentioned earlier this season, the Rockets may have lost their swagger for the time being. And we see it in the shooting numbers as the team feels out the right style of play.
But for now, it looks as if Milwaukee has taken over as the current king of disciplined shooting. As the transformation even converted mid-range king Khris Middleton into a better overall shooter.