Shot Blocking Details: Mining 19 Years of Play-by-Play Data

facebooktwitterreddit
Sep 11, 2015; Springfield, MA, USA; Dikembe Mutombo speaks during the 2015 Naismith Memorial Basketball Hall of Fame Enshrinement Ceremony at Springfield Symphony Hall. Mandatory Credit: David Butler II-USA TODAY Sports
Sep 11, 2015; Springfield, MA, USA; Dikembe Mutombo speaks during the 2015 Naismith Memorial Basketball Hall of Fame Enshrinement Ceremony at Springfield Symphony Hall. Mandatory Credit: David Butler II-USA TODAY Sports /

We’re getting close to having 20 years of play-by-play data publicly available, but the depths of this rich data set haven’t been fully uncovered for basketball fans. If you want some detailed shot blocked information some of the best studies are a decade old — no slight at 82games.com, but we can do better. Taking the cue from Dikembe Mutombo’s recent finger-wagging enshrinement into the Basketball Hall of Fame, and with an assist from Seth for the data, I’ve uncovered a few miscellaneous block stats that aren’t readily available[1. Quick note on the data: due to how it was grabbed, a tiny percentage were not collected, namely ones at the very end of a period. Plus, play-by-play logs do not have 100% accuracy and some games are missing]. A blocked shot is one of the most commonly cited stats and it’s still used today as a measure of a defender’s power, but there’s more to learn when you dissect the stat.

Block Type Leaders

The reason it’s important to slice and study blocked shots is that they are definitely not all equal. Consider trying to grade the best shooters without shot location — isn’t that negligent? You’re mixing up guys who only shoot at the rim, like DeAndre Jordan, with outside shooters like Korver and Novak. Comparing a dunk with a contested fadeaway long two-pointer conjures the old idiom of apples and oranges, or maybe it should apples and dragonfruits.

Blocking a shot at the rim should generally be more valuable than blocking a midrange attempt; few people would argue that. It’s more valuable when your own team grabs the blocked shot too — let’s call that the “Russell.”  Let’s also throw in blocked shots at the end of the shot clock where the offensive team rebounds the swatted attempt[2. Unfortunately, the available data did not have shot clock information, so only situations where the previous event was a possession change (i.e. a field goal attempt from the shot-blocking team or a steal) were counted. This means the lost grenade discussed here is an underestimation. This can be collected in the future.], and give it the name a “lost grenade” as a companion stat to the newly minted “grenade.”

Utilizing regular season data stretching back to the 1997 season, the table below summarizes the findings. The first level of analysis pertains to blocked shot location. Sorting by average distance with a minimum of 300 blocks, there’s a surprising role player at the top: the Vanilla Gorilla, Joel Przybilla. As a Blazer fan, I was actually expecting his name near the top from years of watching him defend. I can’t find a source, but I remember him talking about his shotblocking philosophy. He wasn’t a high-flyer but when elite athletes came inside to dunk on him he knew that no matter how high they jumped, they’d have to come down eventually to the rim, and that’s where he waited. Also, who would have guessed Gerald Wallace, predominantly a perimeter player, would be third in average distance of his blocks?

Looking at dunks blocked per possession in the table below, Alonzo Mourning and Przybilla are in another league with their leads near 50% larger than third place[3. Joel has a higher proportion of blocked dunks to total blocks, however, and Mourning is missing the segment of his career when he had the fewest blocks per possession. As counter-intuitive as it is, that was when he was a young center with Charlotte and one season with Miami. He actually peaked as an older sixth man during the Heat’s Shaq era, blocking a ludicrous 6 to 7 shots per 100 possessions.]. The layup leaders have some similarities with the dunk leaders with Alonzo at the top again and Joel close behind, but there are a few other players adept at blocking layups too like Eddie Griffin and Larry Sanders. Looking at all blocks within five feet of the rim, the mostly forgotten Jim McIlvaine sits on top. He contributed little besides fouling and swatting shots, and it shows. Interestingly, Ibaka is second; his defense has often been questioned by advanced stats but he blocks a high number of shots at the rim. Blocks in the 5 to 10 foot range are compelling. This is where you see blocks on post-ups, so players like Shaq, Brendan Haywood, and Dampier rank higher — you don’t want to post those guys up.

While you can separate blocks into several types, sometimes people combine them with steals without any adjustment, which is a faulty assumption. One main difference between the stats is that steals guarantee the end of a possession, whereas only 57.2% of all blocked shots were recovered by the defense. So which players are the most adept at creating those “Russell” rebounds? Duncan has a huge lead in raw totals, but he has a huge lead in pure blocks too. His ratio (total blocks rebounded by team over total blocks) is above average, but nothing exceptional.

However, a lot of this variation in the ratio has to do with where the blocks are, as the rebound recovery rate changes with respect to shot distance. Looking at the Russell ratio for shots inside of 5 feet, the leader is Jamaal Magloire, whose huge mitts probably helped. Mutombo and Kirilenko ranked well too, while the bottom of the ranking consisted mostly of guards along with a handful of big men like Amare, Hawes, and Shaq. Lastly, shooting fouls committed is a decent proxy for how skilled these shot-blockers are at stopping shot attempts without errors or fouls. Again, the leaderboard at the top has mostly the same set of players like Mourning, Kirilenko, Ben Wallace, Mutombo, Davis, and Duncan, but there’s one surprising name: Eddie Griffin. He had one of the highest block rates in NBA history, and he did this without a large amount of fouls. He died young in tragic circumstances, but he had the potential of a number one overall pick and could never bring his career together.

The Value of the New Block Stats

How important are these miscellaneous block stats? Using a statistical plus/minus model, I can input these new variables into a model and test them alongside a handful of other stats to find their utility. With a few different combinations of variable types and techniques, there are some clear patterns. The first one is that “Russells” are indeed valuable. While it’s no surprise a block-rebound is valuable because the player is ending a possession, there is a question of how much influence a player can have and whether or not it’s luck. At least there’s a pattern — tall, long players usually have high ratios (with the exception of Shaq and Chris Andersen, among others) while perimeter players have the lowest ratios, which makes sense because it’s tougher for them to corral the ball or rebound it. The coefficient is small enough that even with high volume shot blockers the effect is small, but it’s there.

The newly named “lost grenades” were only significant in one iteration of the model and the value was quite low. It’s not a high volume stat, like shot attempts or defensive rebounds, but it’s probably not significantly valuable because a few relatively rare events are picked up by the model, like charges drawn. I was mainly curious about the lost grenade stat because having a shot blocked with five seconds or less is catastrophic for an offense and it could have been a proxy for smart defenders who keep track of the shot clock and know even if the offense recovers the ball, they’re in a desperate situation.

Moving onto shot block location, the most valuable blocks are near the rim. Shots within five feet were consistently found to be significant by the model above the average block, and sometimes shots from 5 to 10 feet were as well. Midrange shots, strangely, also had positive coefficients in a couple iterations of the model, yet blocked 3PTers were only significant in one version and it was with a negative coefficient, meaning it was correlated with worse players[4. I used two different dependent variables: total RAPM and defense only. The 3PT coefficient only showed up in the total RAPM version. This suggests it’s not just correlating with position since perimeter players block more three-pointers and on average have a lower defensive RAPM.]. Looking at the leaderboard again for 3PTers, it’s populated by perimeter players with a few notable defenders, but you also have guys like Harden and Kobe thrown into the mix — they are not defensive stalwarts by any objective measure over the last few years. Perhaps it’s a proxy for defenders who are too aggressive on close-outs and leave their feet or foul. Eddie Jones blocked the most 3PTers with Pau Gasol close behind — I’m not sure what to think there except that long-armed defenders have a higher chance. A more comprehensive investigation is needed here because even though three-pointers are obviously worth one more point, they’re not more valuable for blocked shots.

Breaking things down by shot type, layups, surprisingly, were highly valuable while dunks were assumed to be only as valuable as other block shot types within five feet. I’m still trying to unravel the reasoning behind this, but with layup blocks you have a wider range of defenders including guys who stay on their feet like Duncan whereas to block a dunk you need to really commit and go after the block aggressively. Plus, a layup is easier to control. Of course, another issue that a blocked dunk is fairly rare and over ten times as infrequent as a blocked layup. It’s the kind of event where a season leader generally ranges from 10 to 20 blocks — it’s hard to find a usable signal with so few instances.

Lastly, a simple ratio was very powerful: blocks divided by shooting fouls[5. More accurately, this is blocks/(shooting fouls committed + 1) to protect against cases where a player has no shooting fouls.]. I’ve used the same stat before with personal fouls substituted, and it’s a surprisingly powerful proxy for rim protection and defense in general. However, I’m afraid this is dangerously in overfitting territory because the ratio doesn’t scale well to the per possession form the rest of the model was built to.

Finally, I tested the utility of these new block stats by using a handful of metrics to predict team wins a season later — in other words, a player’s rating in 1997 was used to predict wins for the team the player plays for in 1998[6. Low minute players were regressed to -2, and rookies were all given values of -2.]. I also weighed the errors by the proportion of minutes a team had from new players. This was all done to ensure the additional block variables were useful without being overfit, and for a handful of people who are RAPM non-believers[5. The dependent variable for the model testing to extract the coefficients was 15 year total RAPM].

The models with these miscellaneous block stats were slightly more powerful in predicting team wins in another season — the margins were slim, but they were present. This was all tested on out-of-sample data (seasons 1997 to 2000), and I even tried “backwards” season predictions — using player stats in 1998 to predict team wins in 1997, for example — to see if the results changed, and they largely did not. Additionally, the differences are magnified if you concentrate only on team defensive rating, not just team wins.

Conclusion

With a wealth of data publicly available but hidden in rarely viewed play-by-play files, there’s a great opportunity to unearth a few creative and miscellaneous stats. Some of these might seem unimportant and excessively specific, but we don’t know which stats matter until we evaluate them — and there are always stories to connect to with every one of these stats, from Przybilla’s dunk blocking skills to Eddie Griffin’s overall blocking prowess. There are also a handful of new savants, from Gobert to Noel and from Whiteside to Anthony Davis, who can be analyzed more thoroughly with an expanded set of block stats.

Bonus: Who Has Blocked Dirk’s Patented Fadeaway?

With full play-by-play data the possibilities are an endless horizon. Some tough questions can be systematically answered, like generating a list of players who have blocked Dirk’s seemingly unblockable one-legged fadeaway rainbow jumper. By filtering only for blocked shots on Dirk where he was 8 feet and further out where the shot type was labeled “Turnaround Jump Shot,” “Fadeaway Jumper,” and “Step Back Jump Shot” I have a list of candidates.

Season

Period

Game time

Jump shot type

Shot distance (ft)

Opposing team

Shot blocker

2003

1

0:24

Turnaround

14

NetsRodney Rogers
2004

4

5:36

Turnaround

14

PistonsBen Wallace
2004

1

10:37

Fadeaway

9

SonicsAnsu Sesay
2005

2

6:16

Turnaround

8

SpursRobert Horry
2009

3

0:48

Fadeaway

11

GrizzliesGreg Buckner
2009

2

0:26

Turnaround

12

HornetsDavid West
2009

2

1:50

Step Back

10

NuggetsKenyon Martin
2010

1

6:12

Turnaround

11

RocketsTracy McGrady
2010

3

10:20

Turnaround

12

RaptorsAndrea Bargnani
2011

3

6:38

Fadeaway

11

BlazersLaMarcus Aldridge
2012

2

8:24

Turnaround

11

LakersMatt Barnes
2012

2

4:40

Fadeaway

12

SpursBoris Diaw
2013

2

1:11

Fadeaway

11

LakersEarl Clark
2014

3

10:22

Step Back

13

ClippersBlake Griffin
2014

2

3:39

Fadeaway

12

SunsEric Bledsoe

Unfortunately, there is no information on where the defender was, so it’s possible a few of these are instances where Dirk was blocked from behind, but it’s a good sign a large number of these players are power forwards who would be presumably guarding him. For an example, here’s video of Blake Griffin blocking Dirk’s patented shot with impeccable timing and quick hops. Conversely, here’s an example of a smaller player, Eric Bledsoe, coming over to help and blocking his jump shot.

This question has occasionally arisen in NBA conversations, but hopefully this is a helpful starting point for who’s blocked one of the most famous and unblockable shots in NBA history.