A general finding in the NBA is that jump shots off the catch (Catch and Shoot) generally yield better results than those of the dribble (pull-ups). Several explanations have been put forward for this finding: catch and shoots are usually more open, they happen following ball and player movements that left the defense disorganized, and the motion of catching and shooting is more natural and fluid to most players than pulling up. The truth is probably a mix of all of these explanations, as some causal explorations of the question have concluded that even accounting for defender distance and openness of the shot, C&S are still more effective than pull-ups - but only a little bit, and in much smaller proportions than the raw numbers suggest.
Because of the lack of tracking data, doing this analysis for the WNBA is a much harder task. A potential proxy for whether the shot was a C&S or a pull-up is whether the shot was assisted: a shot that is assisted is either a C&S or a shot following one or two dribbles made for balance purposes. However, we only know if a shot was assisted when the shot was made, we have no information on what came before a missed shot. In statistics, this is called a selection on the dependent variable problem. Here we want to learn about the probability of making a shot (this is our dependent variable), and to learn anything we cannot restrict ourselves to made shots only. Even more problematic, we also do not have tracking data on the defense for the WNBA, which makes any causal exploration of the question such as the one done here for the NBA impossible. We cannot know whether a shot was open (yet), making it impossible to use a matching strategy (compare two very similar shots taken in similar circumstances and facing similar defenses, one being a C&S and one being a pull-up).
So, based on these limitations, can we learn anything for the WNBA? ESPN provides a description for each shot in its data. The descriptions include whether a jump shot was a pull-up shot, a driving or a normal jump shot. For the purposes of this analysis, I assume that all non pull-up or driving shots are C&S. By most definitions, C&S and pull-ups are shots that happen beyond 10 feet, and I follow this convention. I also get rid of outliers (shots from beyond 30 feet). This is not a perfect measurement strategy by any means: it relies on ESPN's manual coding of each shot, and some coders might have different definition than others. The definitions and options for coders also seem to have evolved over the years. For example, pull-ups only appear as a category starting with the 2013 season, and therefore I limit my analysis to the 2013-2021 (up to June 1) seasons. However, in the absence of tracking data for the WNBA, I believe this is the best we can do. As a check, I look at the top 15 players in FGM between 2013 and 2021 for C&S and pull-ups. The most prolific pull-ups shooters appear to be more ball-dominant players and point guards, which is what we would expect, and Courtney Williams dominating the category is a reassuring sign. The best C&S shooters are more pure shooters like Allie Quigley, which is again what we would expect. This is not a perfect check, but it indicates that the ESPN's coding of shots since 2013 provides some information on what I was looking for.
I can now compute the average points per FGA for pull-ups versus C&S. I compute different averages for twos and for threes, because averages across both could just hide different rates of pull-ups and C&S in FGA2 and FGA3 territories. For example, the higher FG% in the pull-up table above could be driven entirely by the fact that a higher proportion of pull-ups are FGA2 compared to C&S. But, as shown by the graph below, even 'controlling' for twos versus threes, the surprising finding sticks: in the WNBA, pull-ups seem to be more efficient shots than C&Ss. A pull-up from 2 has the same expectations for points than a C&S from 3 (1.01 expected point for both).
Based on the best available data, it appears that pull-ups might be better options in the WNBA than they are in the NBA. Points per shot in the NBA for C&S twos are 0.86 (0.66 for the WNBA), and 1.12 for C&S 3s (1.01 for the WNBA). For pull-ups, an attempt from 2 in the NBA is worth 0.79 point (1.01 for the WNBA) and an attempt from 3 is worth 0.97 point (1.1 for the WNBA).
Provided that this data is right (and I don't think it could be that wrong), what are the possible explanations for this finding?
The first thing I do is run a regression controlling for all the possible confounders (i.e. variables that could explain differences between C&S and pull-ups) we have data on: distance and location of the shot, game spread, score differential, minutes on the clock and period number. If pull-ups became harder as the game advance and players become more tired, for example, this could explain part of the differences found in the graph above. Even controlling for these variables, I find that pull-ups increase the likelihood of hitting the shot by 14.9 percentage points (15.2 for threes).
One explanation could be about defense and styles of play: it is possible that C&S are better covered by defenses in the WNBA than they are in the NBA. The NBA offenses have transitioned toward a movement-oriented offense and position-less basketball, which might lead to more open shoots than in the WNBA. Unfortunately, we do not have tracking data for defenses in the WNBA and therefore cannot test this theory.
Another explanation could be that C&S are technically or physically tougher shots to realize for women than they are for men. I don't really believe this explanation, but wanted to mention it and would be curious to hear more opinions.
Or finally, it could be because pull-ups are more rare in the WNBA (because the game is more fluid and the ball circulates more, for example, or because of different shot selection decisions), and only the best pull-up shooters or players in the best situations take those shots. I'm reluctant to rely too much on the count of shots from the data because of my concerns about the quality of the data input and the reliance on manual coding. But based on the data I have, in the WNBA 70% of FGA2 are C&S and 16% are pull-ups. From beyond the three point line, 93% of FGA3 are C&S, 3% are pull-ups. Again, because of how it was coded these numbers very likely underestimate the number of pull-ups and overestimate the number of C&S. Even accounting for data entry errors however, we are still far from NBA numbers: between 2013 and 2019, 32% of FGA3 were pull-ups (and the number is rising), 37.5% were C&S. I am not comparing apples with oranges when comparing NBA and WNBA numbers here, because of significant differences in the data sources, but these numbers suggest that pull-ups are still much rarer in the WNBA than they are in the NBA. This might change with the new generation of players (with players like Courtney Willians already leading the change and new players like Sabrina Ionescu likely to continue the trend). Pull-ups becoming more mainstream might lead to them losing their efficiency advantage in the WNBA.
I am still surprised and slightly confused by these findings. Provided we can trust the quality of data from ESPN on shot descriptions, I would love to hear people''s alternative explanations for why pull-ups appear to be that much more efficient in the W!
Comments