The Houston Rockets defense is crumbling. After being the 8th best defense of last year, with a defensive rating of 103.4, they are down to a 24th this year with a defensive rating of 106.8[1. NBA league average scoring per 100 is more than a point lower than at this point last season – 104.3 this year against 105.6 last year.]. Twitter is full of mentions of Houston’s bad defense. But here is the thing where Charles Barkley might have been right last year: The Rockets’ defense was probably more lucky than good during the last season[2. Of course they way the Chuckster put it didn’t make any sense, but still…]. Looking deeper into the numbers, two things become more and more clear:
As a defender, 3 Point Field Goal % is pure luck
We have written about this
, but it bears repeating this point until the myth finally dies:
The influence a player has on opponent Three Point Percentage is so minuscule that it probably should never be mentioned
. The proof is in
:
If you can see any pattern in here let me know[3. And yes, opponents shot only 30% from three “against” Draymond Green during each the last two seasons, but that is almost certainly randomness in action. Among the sample of all NBA players at least one and possibly more will show up “well” in consecutive years simply by chance.]. And as you don’t see an effect on a player basis, you will also not see an effect on a team basis (as a team consists of players…). New York was last in Opponent 3FG% last year and leads the league at the moment. The is the result of statistical randomness far more than Kristaps Porzingis blocking a jumper or three. So, while Houston “led” the league in opponent three point percentage (32.2%) during the 2014/15 season, they have fallen to the middle of the pack (34.1%) during this season. With around 25 three point attempts per 100 possessions, that is a 1.5 point drop in defensive rating right there.
Points per possession allowed on transition plays is luck
Thanks to nba.com’s playtype data, we can see how much opponent teams scored on transition plays during last season and this season. The result is similar to 3 point percentage:
As you can see, there is no rhyme or reason for why you score more or less in transition against a certain team. Sure, there are players that make the highlight chase down block, but these highlights happen maybe once every 3 games, while there are around 15 transition plays per game. Once again the noise outweighs by far the signal. As you can see here, Houston allowed slightly below average points per possession last year and above average this year. The difference is around 0.08 points per possession, which is especially important for Houston, as:
Houston allows a ton of points in transition per game, something that is clearly controllable, not by defending “transition plays” better, but rather by not allowing them in the first place.[4. It is probably not advantageous to have two shooters on the baseline and the ball handler driving. But this is not part of this article…]. So, that is close to another 1.5 points drop in defensive rating, which together with the difference in three point percentage gets you close to the 3.4 drop in defensive rating that they actually have.
As an further note: These luck factors (and probably more) are most likely a big component in all current plus/minus stats including adjusted plus/minus models such as RPM and RAPM. More on this to hopefully follow soon.
Why I think that Houston’s defense is actually not so different from last year
Going back to my idea of distancology, it is possible to deduct a lot of information about a team by looking at shot distributions. In this case I clustered all teams for this and the last season after opponent shot distance frequency and added information about field goal percentage. So in the left half of the figure blue means a low shot frequency and in the right half blue means a low field goal percentage.
This graph does not show information about offensive rebounds, steals and free throws, but it shows some of the most important things for a good defense: 1. Protect the rim and don’t allow shots between 0 and 4 feet) (two blue columns on the left) 2. Avoid three pointers (two blue columns on the right side of the dashed line) 3. reduce the field goal percentage close to the rim (3 blue columns on the left of the right half[5. Sorry that I can not explain this better].
A lot of other things are out of the defense’s control, most importantly opponents’ three point field goal percentage. But you can see that two off the best two defenses of this and last season, the Spurs and the Warriors, are for both years on the top of this graph, which is the place to be.
Houston on the other hand falls exactly into the same cluster where they have been last year, allowing too many shots at the rim and corner threes. Only this year, Lady Luck has not been as good of a help defender as last year.