During the NBA’s Christmas quintuple-header, the league required all of its players to wear sleeved jerseys. With the league expecting to add more games with sleeved jerseys in future years, the rush to analyze the effects of the uniform change were swift.
Players’ reactions to the uniform change were mostly neutral or negative. Knicks guard Beno Udrih, for example, insinuated the switch caused him to shoot worse, stating “Personally it bothered me and my shot.”
Others took a more analytic approach.
To asses how uniform changes affected shooting outcomes, however, there are probably better ways than simply comparing teams’ overall field goal percentages. First, field goal percentage is largely impacted by opponent and game location, each of which is held constant on Christmas Day, but varies over the course of the player’s career. Second, even within a team, not all players actually play in a Christmas Day game.
Thus, in analyzing players’ shooting, I thought it was important to hold player, opponent, and game location constant. I started by extracting all Christmas Day players who attempted more than 10 field goals (while wearing sleeves, of course). Next, for each player, I used basketball-reference.com’s game-finder to isolate games played by each player which were both (i) against their Christmas opponent and (ii) either home or away (the same as in the Christmas Day game). This, of course, is to compare apples to apples, oranges to oranges, and shooting percentages against the Bulls at home to shooting percentages against the Bulls at home.
For example, Miami’s Chris Bosh had shot 46 of 99 (46.5%) from the field on the road against the Lakers entering Wednesday, and he finished 9 of 18 on Christmas while wearing sleeves (50%). These were the two percentages which I thought were the best to contrast.
Below lies a plot of each player’s shooting percentages entering the game (opponent and location specific, without sleeves), and their shooting percentage on Christmas (wearing sleeves). Players are arranged in order of FG % entering the game.
While there is obviously some variability to be expected in the small sample size of just one game, the majority of players shot relatively close to the individual opponent, location, and non sleeve-wearing averages.
A few players, however, deviated from their expected averages. Serge Ibaka and Dwayne Wade, for example, shot more than 20% better while wearing sleeves, relative to their previous averages shooting at New York and at the LA Lakers, respectively.
While Udrih didn’t meet my requirement of attempting at least 10 field goals, Harrison Barnes and Andray Blatche were two players that shot noticeably worse with the sleeved uniforms, as both shot roughly 30% worse with sleeves on than in games which they played at home against the Clippers and Bulls, respectively.
Of course, with so many players and such small sample sizes of players wearing sleeves, we’d expect a few players to under or outperform their unsleeved expectations. As a result, it’s difficult to tell if Barnes and Blatche are uncomfortable wearing sleeves or simply had off-nights. Further, for all players, there could be an effect of playing on Christmas (due to the holiday itself or the different game times) which cannot be accounted for until teams wear sleeves on different days.
Lastly, while I’d have preferred to have used both a Bayesian model and more player-specific data, I modeled field goal shooting percentages depicted above using a longitudinal linear mixed effects model, with random effects for each player and a fixed effect for jersey type (no sleeves, sleeves). Not surprisingly, no jersey effect was evident (p-value = 0.90).
At this point, the jury is still out on the effects of wearing sleeved jerseys. Early returns, however, suggest that there is no strong impact of wearing sleeves on field goal shooting.
great analysis–i do have one comment though. I certainly understand the benefits of using the opponent-specific, location-specific data, but I do think there’s potentially an unmentioned limitation, needing to extract the data from these specific games means that most come from past seasons…my thought is that players’ FG% may change (possibly predictably, you may even have a better idea of whether it goes up or down) over a long career. Using only recent data would almost certainly not be worth the loss of your apples-to-apples approach, just hinting that if, for example, most players shoot worse as their career goes on, that we might miss some of the effect of the sleeves. (although personally i think it’s just the predictable whining that ensues after missing shots.)
Hi Dan, thanks for reading.
I’d 100% agree that this analysis is far from complete, as, among other weaknesses, shooter age is not consistent in both sleeveless and sleeved measurements. Definitely something to keep in mind if I look further at these numbers.