One of my “quirks” as editor of Wages of Wins is to spoil the surprise. Kevin Draper (@TheDissNBA) of the Diss has graciously added another guest post. This time he’s discussing some of the issues with Assists, and shows what happens when we factor in pace, team ability and three point shooting. The spoiler? These things don’t really add much to the vanilla assists per game!
If you are reading the Wages of Wins, you are probably aware that many box score statistics have major limitations. For instance, points scored isn’t the best measure of whether a player is a good scorer or not, as Dre’s post on Kobe Bryant illustrates. Perhaps the most misleading of all box score statistics is the assist, which suffers from numerous problems:
- Courtside statisticians inconsistently employ the definition of an assist when crediting one
- Assists are a counting statistic, meaning they are somewhat a function of playing time, not ability
- Assists do not account for pace
- Assists do not account for the difference in value of two- and three-point shots
- Assists do not account for teammates’ shooting ability
The first problem is impossible to correct fully, unless somebody comes up with a computer program to calculate assists, but the other four can be partially, if not fully, mitigated.
2012-13 Assists per 36 Minutes and per Game Leaders
The above table shows the importance in controlling for minutes played. If you only looked at the leaders in assists per game, you would miss out on the fact that players like Jamaal Tinsley, Andre Miller and JJ Barea create a lot of good shots for their teammates, but don’t make the assists per game leader board because they only get limited playing time. Simply looking at assists per 36 minutes (AST/36) instead of assists per game (AST/G) is the best, and easiest, way to have a better understanding of how well a player creates for others. Let’s see what happens when we tweak the data further.
To account for the differences in pace, I employed the equation (AST/36 * average pace / team pace). This increased AST/36 for players on slow teams, and decreases AST/36 for players on fast teams.
To account for teammates shooting ability, I employed the equation (AST/36 * average team FG% / team FG%). This increased AST/36 for players with poor shooting teammates, and decreases AST/36 for players with good shooting teammates.
Three-point Shooting Adjustment
To account for the added value of assisting three-point shots, I employed the equation [AST/36 * (1.5 * (total team 3 pointers / total team field goals) + (1 - (total team 3 pointers / total team field goals)))]. This increased AST/36 for all players (assuming their team has made at least one three-point shot this season), but did so more for players on teams that make a lot of three-pointers.
To make the new adjusted assists per 36 minutes (ADJAST/36) total comparable to AST/36, I then multiplied ADJAST/36 by the equation (sum of all players’ AST/36 / sum of all players ADJAST/36).
There are a couple serious limitations to my methods. None of the adjustments accounts for players’ on/off splits. It is very possible that, say, the Knicks play faster when Raymond Felton is on the court than Jason Kidd, and also quite likely that the Clippers shoot better when Chris Paul is on the court than Eric Bledsoe. Additionally, the shooting and three-point shooting adjustments include players’ own contribution to those team statistics, even though it is impossible for a player to assist his own basket. The tradeoff to avoid these limitations is calculating individual adjustments for each and every player; a tedious task that, as will be shown shortly, would have very little benefit.
2012-13 Adjusted Assists per 36 Minutes Leaders
The above table shows how little pace, shooting, and three-point shooting adjustments changed assist totals. The biggest difference the entire sample belongs to DJ Augustin, who went from 6.5 AST/36 to 7.2 ADJAST/36. The Pacers play at a very slow pace, don’t shoot very well, and make an average number of three-point shots.
When we think about the reason there is only a slight change in the numbers, the answer makes sense. The difference between the fastest and slowest teams accounts for about six extra possessions per 36 minutes. Given a league average FG% of 44.6%, this is about three extra made baskets per 36 minutes. The difference between the best shooting team and worst shooting team accounts for five extra made baskets per 36 minutes. The difference between the heaviest three-point shooting team and lightest three-point shooting team accounts for six extra made three-point shots per 36 minutes. Given that the three-point shooting adjustment adds half an assist for a made three-pointer, this is three potential assists per 36 minutes.
A player on the slowest, worst shooting, fewest three-point shot making team has 11 fewer chances per 36 minutes to have gained an assist. Compared to a player on an averaged paced, average shooting, average three-point shot making team however, that same player only has 5.5 fewer chances to gain an assist. In reality though each team has a distinct mixture of pace, shooting ability and three-point makes. Therefore, the assist adjustments only corrects for a couple extra assist chances per game.
The above graph shows that the relationship between AST/36 and ADJAST/36 is practically perfect. The R2 value indicates that virtually all variation in ADJAST/36 is explained by AST/36. Or, to put it in another, slightly incorrect way, the three adjustments made less than a one percent difference. They did almost nothing to help better understand who the best creators in the NBA are.
If you want to know who the best assist-men in the NBA are, just go look at the assists per 36 minutes statistics on any website that has them. Of course, ignore the players who have only played a few minutes this season.