Rankings are among the most popular exercises for most NBA fans, and also among the least efficient ways to evaluate data. The enduring appeal of rankings and lists owes to their relationship, however tangential, with the big questions: Who is the best? Which player is better? How much better or worse would a team be if they replaced this player with that one? Every fan, analyst, scout, coach, and executive needs to be able to answer these questions.
The problem with rankings is not the questions they address, but the biases implicit in ordinal numbering. Our brains are trained to think that the difference between 12 and 17 is the same as the difference between 18 and 23, and that all these values are on a different order of magnitude from single digit numbers. When we transfer these assumptions to rankings of the “Top 50 Players in the NBA” or something similar, we start with false presuppositions. There is not an identical difference in value between each pair of contiguous players on any player ranking. Even a hypothetically perfect, pie-in-the-sky ranking that ranked players in the exactly correct order would still need something besides the ranking to indicate the players’ true value relative to their peers. Our habit of using ordinal numbers to rank players blinds us to the shape of the data.
As a result, both league insiders and outsiders spend an inordinate amount of time debating questions such as “Is Kevin Durant the second-best player in the league or the eighth-best?” “Is Paul George a top 5 player, a top 10 player, or a top 15 player?” What really matters is how much a player contributes to wins by helping put points on the board (on offense) and keep points off the board (on defense). If the 6th ranked player in the league and the 16th ranked player are nearly identical in terms of their contribution to team success, it makes little sense to lay so much weight on their difference in the rankings.
What a valuable ranking system can tell us is how much value a player generates, relative to the rest of the population. In answering the question posed in the title of this article, I will attempt to provide enough context for the reader to be able to comprehend the shape of the data. As such, let’s start with refining the question itself:
Who are the best defensive point guards in the league?
When we ask these types of questions and address them with data (whether we’re using statistical data, scouting data, or even anecdotal evidence), we are not asking, “which player has unusual data associated with his performance in my dataset?” We want to know which players our data says contribute the most to winning. When we evaluate defense, that contribution takes the shape of preventing the opposition from scoring.
Right away, this definition reveals an asynchronous relationship between players in the same position group. In order to prevent the opposition from scoring as effectively as possible, a team needs to allocate their greatest assets to preventing the opponent’s best scorer from putting up points. While the “payoff” may not be immediately obvious in the player’s individual statistical record, it is clear that the team’s fortunes turn on the ability of its stopper to slow down the other team’s star.
The Effect of Defensive Load on Per-Possession Defensive Performance
If the team needs to stop the opposing star, then the team’s most talented defenders are often going to be given the most difficult assignments. The best defender is the defender who can carry this heavy load on defense while keeping as many points as possible off the board. My Matchup-Based Defense model accounts for this necessity by comparing a player’s effectiveness in shutting down his man with the load he carries for his team.
The best defensive point guards, from this point of view, would be the players who carry a heavy load for their team on the defensive end, but who manage to be effective against even the best scorers in the game. Last season, there were 17 point guards who carried a high load for their team. These players faced either a) a raw load of 13 points per game, or b) a relative load at least 5% above average on their team. These 17 players exhibited a noticeable range in effectiveness.
In the chart below, every player whose average position (using basketball-reference.com’s height-based positions) was less than 1.5 is placed in one of four groups. The first group is the high-load defenders, whom we’ve just mentioned. The second group, labeled medium-load defenders, are players who play a lot of minutes, but are not crucial to their team’s defense. They sometimes have to defend scorers, but their coaches would rather give them an easier defensive assignment when it is feasible to do so. These players face either a raw load of 12.0 points per game or greater, or else a team-relative load higher than 1.0.
The third group – role players – is comprised of players who must have a narrowly defined defensive role in order to succeed. These players are typically limited in some way (height, quickness, ability, experience) that either limits their court time or severely constrains their teams in putting lineups around them. Defensive role players need to be “hidden,” much like offensive role players. For our purposes, role players are defined as player carrying a load between 10-12 opponent points per game.
The fourth group on the chart is labeled “small load” due to the stark difference in opponent talent faced by these defenders as compared with their positional counterparts. The small load defenders all face opponents with weighted averages below 10 points per game. As you might guess, many of the players in the small load group are benchwarmers and fringe guys, who defend only other bench players.
There are several interesting notes about the population that I’d like to mention before moving on. The high-load defenders are the smallest group, but exhibit greater variance than the medium load defenders and the role players. Since the chart examines per-possession performance, and since the small load group contains a number of players with a small sample size (meaning that they did not play very many minutes last season), the small load group has the greatest average absolute deviation.
The gray group – the role players – is the most populous of the groups, with 36 players meeting the criteria. Despite having more data points, the group has significantly less spread than the other groups. These players’ per-possession performance is very similar despite variance in minutes played; one explanation for this observed consistency would be that they are, in fact, fulfilling similar roles on defense.
Finally, note that the lower quartile of the high load defenders (in this case, four players) compares favorably with the third quartiles of the other groups. Carrying a heavy load doubtless impacts a player’s effectiveness, even when we adjust for such an effect as in Matchup-Based Defense. Fortunately, even comparatively overworked stoppers are still more effective than many of their position mates.
Who are the Best High-Load Defensive Point Guards?
Now, let’s deal with that group of high-load defenders. Logically and empirically, we can expect to find the best defensive point guards among their number. Who are the best defenders of the group? Seven of the players in our group saved at least 10 points per 100 possessions. Ranked by per-minute effectiveness, then, the best defensive point guards in the NBA last season were:
- Frank Ntilikina
- Derrick White
- Darren Collison
- Ricky Rubio
- John Wall
- Kris Dunn
- Cory Joseph
The tale below allows you to go deeper than a simple list, as promised. You can sort entire sample of high-load point guards by total defense, per-possession defense, defensive load, and team-adjusted defensive load. The reader will also notice that some players often labeled as point guards do not appear in the table below, and will be evaluated with the shooting guards in the second installment of this series. So if you’re thinking “Why aren’t Marcus Smart or Jrue Holiday one of the top defensive point guards?”, the answer is that they are grouped with the SGs.
High Load Defensive PGs
|PLAYER||Position||Total Defense||PER 100 Poss||LOAD||RELATIVE LOAD|
Who stands out in the data? Two players who have earned All-Defensive honors in their careers, Eric Bledsoe and Patrick Beverley, fall just a bit below the top group. Although they “only” saved 8.3 and 8.2 points per 100 possessions, respectively, they did so against vastly stronger competition. Both Bledsoe and Beverley had to shoulder a staggering load in their team’s defensive game plans, and were slightly less effective per possession than a handful of other stoppers. Still, if we were front-office executives interested in acquiring the best defender, it might make sense to target a player who is overburdened in his current organization. After all, it would be perfectly reasonable to expect players such as Beverley and Bledsoe to be even more effective on a team with greater defensive depth.
Derrick White was a breakout performer last year, and did receive 11 votes for the All-Defensive teams. San Antonio’s defense took a major hit after the Spurs sent out Kawhi Leonard and Danny Green in the blockbuster trade with Toronto, then lost Dejounte Murray to injury before the season began. The lone bright spot was White, who saved the Spurs 12.2 points per 100 possessions according to the Matchup-Based Defense model.
Though slightly less proficient in creating turnovers than other top defensive point guards, Kris Dunn more than made up for it by locking down his opponents. Dunn recorded his lowest steal rate as a pro last season, and other metrics assigned him a minimal defensive value. By employing matchup-based data, however, we are able to recognize that Dunn continued to exert a substantial effect on the players he defended, despite recording fewer steals than normal on a bad defensive team. Using other current defensive metrics would likely understate Dunn’s value, since other models (RAPM, WOWY, RAPTOR, DRAYMOND, LMNOP … that last one was a joke) do not directly base their valuation of a player’s performance on the caliber of players he is assigned to defend.
Frank Ntilikina rates surprisingly high due to his preternatural peskiness in forcing turnovers. As I found in testing the hypothesis that good defenders may force turnovers without getting credit for the steals, Ntilikina’s opponents turned the ball over at a stupendous rate in games that the Knicks recorded 10 or more steals. Ntilikina’s pressure on primary ball handlers often resulted in steals for his teammates, helping to explain his high value in Matchup-Based Defense.
Which players surprised you? Let us know in the comments, then be sure to come back by for the top defensive shooting guards in the league!