The Unexpected Defensive Power of the Cleveland Cavaliers

If LeBron James were to win a title with a group of all-star-less role players, one would not have guessed it would have been with this crew. A successful formula for a centralized team like this would involve an elite defense, which is something the Cleveland Cavaliers could not have called themselves even after the trade for Timofey Mozgov. However, Cleveland has been limiting teams and grinding games to a halt like it’s the 2005 Finals. But how much of this is sustainable? How much credit should their defense be given for opponent field-goal percentage?

Using the SportVU shotlog data, you can build a model that estimates the chances of a shot going in based on shot distance, the time left on the shot clock/game clock, defender distance, whether or not it’s a catch-and-shoot attempt, and the individual player’s average on jump shots. Of course, this still misses a lot of stuff like if the defender is between the shooter and the basket and how stationary the shooter is, but it’s a huge improvement over the regular FG% people are used to. Using two years of data in the regular season where the 2015 season was weighed twice as heavily, I built the below model using nonlinear regression in R for shots from 10 feet to 30 feet.

( 74.4 – 1,32 * ShotDist  + PlayerAverage)
[1 + exp( -( 0.345 * DefDist + 0.268 * C&S – 0.00165 * (24-SecsLeft)^2 ) )]

The “PlayerAverage” is centered at 0 where the best shooters, like Stephen Curry, generally have a positive coefficient far above 0 and the worst shooters have a negative one. From that, I can calculate expected points on shots within 10 to 30 feet.

For the Finals, the Golden State Warriors have a disparity of 17.4 points between their expected net total and their actual points scored on those shots. Since the series has been separated by 0.3 points on average, that is highly significant, even if you factor in the loss in efficiency from fewer offensive rebounds. (The Warriors, oddly enough, have had a higher OREB% than the Cavaliers.) But this cuts both ways: the Cavaliers have a disparity of 11.1 points. Both teams have actually shot worse than you’d expect, which should be no surprise in such an inefficient series.

Going game by game, the Warriors had a difference of -4.4 points and the Cavaliers 9.8 for the first game. This means the Warriors shot better than expected, and in the real world the Warriors won by 8 points in overtime. If you look at shots before overtime, the differences change to -6.7 and 4.7, respectively, for Golden State and Cleveland. Based on each player’s shot history, the Warriors were lucky to escape with a win. However, in the second game, the Warriors had a difference of 17.0 points — huge for just jump shots — and 5.9 points for the Cavaliers. That’s the game where you can state the Warriors lost because the shots they normally make wouldn’t go in. Finally, in Game 3, the Warriors had a difference of 4.8 points and the Cavs -4.7 points. Cleveland ultimately won the game by five points, and at the very least it should have been a close game, perhaps another overtime.

Putting it all together, the Warriors should probably be up 2-1 based on their historic averages, but they were actually lucky in Game 1, and Game 3 should have been close no matter what. If you use a conservative estimate that 50% of the point disparity is “real” because of offensive rebounds and other things the defense is doing that isn’t picked up in the model, the Warriors should be leading by an average of 3.5 points per game where they win Game 2 outright and the other two games are virtually tied. The Warriors aren’t just failing because of their jump shots, but in the finals every little bit helps, as LeBron is a one-man elite team and any NBA squad can steal a game from even the league’s champion. The Cavaliers are doing extremely well in limiting free throws and transition points.

Looking at the expected points results for the Golden State Roster, Curry has definitely under-performed but he’s not the biggest culprit. Both Barnes and Green has shot drastically below their historic norms, and they’re dragging down the team almost on their own.

Table: Expected points on 10-30 shot attempts compared to actual points, GSW

Turning to Cleveland, LeBron should be shooting a little better, but the difference isn’t huge. His shots have just been very difficult. There are no standouts on this table, however. Smith and Shumpert could shoot better, but James Jones has been shooting very well.

Table: Expected points on 10-30 shot attempts compared to actual points, CLE

Traditional defensive metrics look at how teams allow points, and sometimes where those points come from. But teams have little control over the FG% of open shots, and they don’t have total control over semi-contested shots either. We can look at how teams score and what their historic averages have been for a complete picture of what teams are doing. The Cavaliers most likely won’t keep holding the Warriors ten points under their league average on efficiency, but they deserve some of that credit. The there’s the issue of when to trust the recent data more, as Draymond does not look like himself and his regular season averages may not apply. It’s up to our interpretation and the application of these advanced stats to figure out what their share of the credit it and what is truly luck and what’s real.