Welcome to another edition of The Professors Corner…and we should probably start with a wellness check. How are you good folks doing? Are you striving for balance in your life? Drinking plenty of water? Stretching? (you’re only as young as your spine) Good. Okay, being adults about this…let’s go.
Ripping off the band-aid— In the last 15 games the Spurs have:
- Won 3 games.
- Produced an offensive rating greater than league average 4 times.
- Shot above league average in true shooting 4 times.
Did you survive? Of course you did! Why? Because sports are fun! And, if you’re like me, you still sincerely enjoy watching the San Antonio Spurs play basketball. As you can see by the Season Recap Chart it hasn’t been easy since the Utah game on December 27th. The Spurs lost Dejounte Murray on December 26th and dropped 4 of the next 5 games. And, once that was complete, the Spurs lost Derrick White and dropped the next 5 games. This was also noticeable beyond the win/loss column since 4 of the 5 lowest assist percentage games of the season occurred in the most recent two weeks. Fortunately, D&D completed their H&S dungeon crawl and reunited to earn a win against the LA Clippers (sans $78.7 million in payroll) and the OKC Thunder.
I noted a positive trend on the right side of the chart above. The Spurs are gradually pushing their offensive rating upward and there are some granular details that support this subtle ascension. Keldon Johnson has returned from Health and Safety protocols with a very impressive true shooting of ~65%. In addition, Dejounte Murray also has rolling 5-game true shooting average that is comparable to his season high from late October. Finally, if you squint, with the sun at a perfect angle, you can see Mr. Bates-Diop hovering right above league average in the same variable.
As I close this introduction, I feel it’s worth noting the Health & Safety protocols were particularly rough on the Spurs. Looking across the different advanced metrics (EPM, LEBRON, DRIP etc...) you would see the club has three players that would start on most teams: Derrick, Dejounte, and Jakob. Two of those were lost in the last 3 weeks and we paid the iron price for that loss. All three of those guys are good basketball players. In fact, one of those two players, Dejounte Murray, is one of the league leaders in triple doubles and the inspiration for this analysis. Thus…
Find an open headspace, grab a brush, grab bucket of indoor satin, and let’s paint ourselves into: “The Professor’s Corner.”
The Triple Double
First, let’s get this straight, the Triple Double is dumb. Don’t get me wrong, it’s a cool accomplishment, but it’s the epitome of a superficial “stat.” The “stat” exists nearly on the sole basis that adding an extra digit to the left of the decimal is super neat. Unfortunately, just by asking some commonsense questions the cracks in the foundation appear.
For example, are 10 rebounds significantly more valuable than 9? How about 10 vs 9 assists? How about points…never mind. Digging further, does the player that averages 8.6 assists per game (vs. 10.2 assists) make significantly less passes per contest that contribute to his/her teammates scoring more optimally? Does the center that corrals 11 rebounds per game (vs one that rakes in 9.5) increase the offensive or defensive efficiency of an NBA team? Are there multiple regression models that suggest crossing from single to double digits in a variable contribute more significantly to winning versus the crossing of other “thresholds”? Those are important questions, most of which are beyond the scope of this piece, but I list them to illustrate the types of questions people should ask about common sport statistics. In this piece I want to focus on the connection between the triple double and winning. To start let’s establish a baseline understanding of triple doubles over the last 35 years of the NBA.
Frequency and Occurrence
Triple doubles are rare, but their infrequency is not equally distributed across the last 30-40 years. For example, the 1980s produced more triple doubles, non-overtime games (n = 459) than the 1990s (n = 324) and this is primarily due to the change in pace. The NBA experienced a significant decline in pace following the late 1980 and early 1990s. For reference, the post-merger carefree years of 1983-86 produced 101-103 possessions per 48 minutes dropping to an entrenched 88.9 possession in the 1998—1999 season. If you examine the blue line on the right side of the chart below you can see the drop in triple doubles coincide with the decrease in pace (light to dark change in color of circles) on the chart. The pace stayed low for nearly 20 years, but eventually rule changes, some of which encouraged the playing of actual defense, and the subsequent changes in offensive strategies to challenge said defenses, resulted in distinct changes in pace around ’13-‘14 season onward.
The 2015-2016 season produced the onset of the cambrian explosion of triple doubles. From ’14-’15 to ‘16-’17 the count increased from 46 to 112 and raised even further to it’s peak in the ’20 –’21 season of at 139 triple doubles. To put this into perspective, there have been more triple doubles in the last 5 complete seasons (580) then there were in the previous 16 seasons (567).
What is very notable about this eruption in triple doubles is that drastically exceeds the increase in pace of play seen in the league. There are other factors afoot and the most important is the augmented employment of what the smart people call more “heliocentric” offenses and I call, “give the ball to your best player, like, a lot” approach to basketball. For example, just brainstorming here, imagine a 6’5” guy with a big beard; or 6’4” that’s an atomic athlete, a 6’7” Slovenian with a frightening step-back 3-pointer, a 6’8” built like Colossus from the X-men, or 7’0” Serbian with the ability to make every pass and you give him the ball to control the offense EVERY TIME. As a Spurs fan, try to imagine Manu, at the top of the key, for the last 2:27 of the games in 2009…but for 36 minutes. Turns out this produces of piles of Triple Doubles! (ps Manu never got a triple double). But lots of numbers isn’t the goal. The goal is to win games.
As a basketball fan this explosion in the frequency of triple doubles caused my eyebrows to raise. Listening to the announcer rave about a triple double that occurred was once every 40 games felt exciting. But with the frequency increasing from 1 every 30 games to the present 1 every 10 games diluted my enthusiasm. In addition, I became aware a foundational assumption built into my understanding of basketball—triple doubles meant winning. If somebody played great enough to achieve a triple double that meant they likely led their team to victory. And it’s with winning that this chart gets really fun.
This chart displays triple doubles, year-by-year, since the 1988-1989 season (overtime games excluded; playoff games included). The blue line, which we’ve discussed, displays the number of triple doubles each season.(1) The circle coordinates are aligned with the average winning percentage of the teams that had a player earn a triple double. For example, in the season ending in 2001 there were 43 triple doubles and the teams with the players that earned those triple doubles won 33 of those games and lost 10—thus a winning percentage of 76.7%. For added information, the circle color represents the average pace of play for that specific season. Finally, circle size displays the frequency of triple doubles per season (just like the blue line) to visually reinforce how this variable has changed over time.
On this chart I also overlayed the line of best fit and 95% confidence intervals produced from a logistic (binary) regression. This analysis compares the occurrence of wins/losses over time and calculates a mean win probability and subsequent confidence intervals. For example, in 2005, a triple double had a win probability of 77% with a surrounding confidence interval of 74-80%.
A decline in Winning
The location of the colorful circles on the chart represents the winning percentage of the teams for which the player that earned the triple double played. As you can see, it’s declining. Via the 95% confidence intervals the winning percentage ranged from 78-88% in the early 1990s. This range has declined to 66-76%, but more importantly, the explosion in higher frequency has created a tight cluster, on the far right of the chart, orbiting around 71%— with a nadir in winning percentage produced by the current season of 62%.
Triple doubles are cool, but if they come attached with a winning percentage of 62% then I’m much less enthusiastic. It’s not a straight 50/50, but it’s also not a hefty 85% either. Celebrated statistics should show a quantifiable value to winning and a winning percentage of 62% is too low for me. But, with that stated, 62% is likely the low end of the winning percentage resulting from this heliocentric-induced frequency of triple doubles. I doubt we will see many seasons lower. I say this because the occurrence of triple doubles are still rare enough that certain teams/individuals can “pull” the statistic artificially in a direction. Triple doubles are not evenly distributed across the league. There are a small group of guys that account for most. Currently those folks have names like Harden, Jokic, Westbrook, and Murray and three of those are on teams have records hovering around .500 or below. But, if another individual, say Luka, shreds the second half of the season, and accumulates lots of wins and a tasty assortment of triple doubles, then the average winning percentage across the league may climb back up to 70%. But, again, I do feel 60% is extremely low and let’s address why.
Assists: The rate-limiting enzyme of the Triple Double.
In every process there’s a step that “governs” the process. In physiology, the slowest enzyme of a metabolic pathway is called the rate-limiting enzyme. In the triple double, that enzyme is the assist. Via nearly 6,000 box scores from this season (>24:00 minutes played (1)) triple doubles (again, non-overtime) occurred in ~1% of those performances. It’s important to note the variables that produce a triple double occur in different frequencies. Of those that earn 24 minutes of playing time, roughly 76 % of those will score 10 or more points. This is logical—three 3-pointers and a free throw will get you into the club. Earning 10 rebounds occurs in 1/5th of the that (14.7%). But way down at the bottom is 10 assists at only ~4% of individuals performances with at least 24 minutes on the court. Assists are so crucial that ~1/5 of the performances that reach 10 assists achieve a triple double.
We’re here to talk about winning and that’s where this is becomes more interesting. I had this hunch: For a player to produce a lot of assists his teammates need to make shots (total galaxy brain). This can be supported by the fun fact that all top 10 of the current assist leaders in the NBA are on teams that shoot over the league average in field goals percentage. But let’s look closer.
I examined the triple doubles since the fall of 2018 (n = 417 triple double games) and cross-checked those with the basic field goal percentage for their TEAM in that game. Then, this is crucial, calculated the LEAGUE field goal percentage for the month in which that game occurred (the mean league FG% changes through a season) and calculated the difference. Using the monthly league average for field goal percentage is important since the empty gyms of the pandemic inflated shooting accuracy. For example, the “empty gym” season produced a mean field goal percentage of 46.6% while the current season has a mean of 45.3% (the mean across the three most recent 3.5 seasons was 46.1% . Finally, I did a statistical test that asked, “Are the field goals percentages in triple double games significantly better than league average?”
Absolutely!(p= 0.001). In those 417 games that players earned a triple double their respective teams averaged a field goal percent of 48.5% — and the mean was 2.4% better than the league average. That doesn’t seem like much, but please consider that the range of team averages is spread from OKC at 41.7% to the Suns at 47.6%— a very tight 5.9%. Therefore, teams in triple double games averaging 48.5% is significant since it exceeds any current team average. Consequently, (hint: your ears should be perked up), wouldn’t teams that shoot above the league average, regardless of a triple double, win a decent percentage of games?
Heck yeah they do! Here is a list of 7 of the top 9 teams in field goal percentage this season: Suns, Bulls, Jazz, Nets, Nuggets, Warriors and Cavs. See a trend there? For further evidence, early this season I posted a tweet from a regression analysis in which I examined the win probability based upon the difference in true shooting between two teams in a game. A 2% edge in True Shooting produces a 70% win probability and a different of 6% raises the probability 90%! (+/- confidence interval of 9%) Finally, looking at the winners of the 658 games played this season, 455 (69.1%) of the winners shot above league average in field goal percentage while 203 (30.9%) shot below. Which is extremely close to the average triple double win percentage of the most recent 5 seasons. And that, my dear friend, brings us to a conundrum of causality. Let’s review:
- Triple doubles of that last several seasons, aka “the heliocentric era,” are associated with a ~70% winning percentage.
- The limiting factor for most triple doubles is obtaining 10 assists.
- Assists require a pass and a made shot and most (~68.1%) triple doubles occur when the team shoots above league average accuracy.
- Shooting above league average accuracy is ALSO associated with winning percentage of ~69%.
What are we celebrating?
It’s very difficult to isolate if a triple double in this heliocentric era is the result of a great individual performance or above average shooting by a team that night. If a player governs the ball on most offensive possessions, and his teammates shoot accurately, plenty of assists will occur and that team will likely win the game. Thus, there’s a decent argument that the team gets the triple double for the player and that we should shift our celebratory focus to support the team instead of the individual. For example, recently, against OKC, Dejounte Murray had 14 assists and Doug McDermott also hit 6 of 9 3-pointers. All 6 of those 3-pointers were from Dejounte’s passes! Therefore 6 of 14 assists involved an awesome shooting performance.
But there’s plenty to support the individual achieving the triple double. The most obvious is that a player must be extremely good to be trusted with the ball. Most teams only have a couple viable options. If you gave an average NBA player a very high offensive load they might get a triple double… but it would involve turnovers. The other, arguably more important reason is that there are certain players that CREATE great scoring opportunities for their teammates. And those opportunities are associated with higher quality shots that are more likely made than missed—therefore increasing field goal percentage of the team. There are quantifiable ways to investigate this but the purpose here is to establish a foundation to understand the modern NBA and his long revered statistic.
My personal conclusion is that this heliocentric era has forced me to re-calibrate my opinion of the triple double. I now see it more often as a product of a solid team performance that manifested a notable statistical accomplishment by a VERY important member of the team. I also speculate, moving forward, that the winning percentage of this accomplishment will average slightly below 70%. Which is really good! A team winning 70% of games is a contender. Thus, I’m closing this Professor’s Corner feeling positive. I like teams and I like teamwork. I feel a triple double implies a team trusted the right guy to have the ball and trusted each other to shine when that player delivered them the moment.
Notes: 1) Russell Westbrook is the only player since the 88-89 season to earn a triple double while playing less than 24 minutes. On March 3rd, 2014 he had 13 pts, 10 reb, and 14 assists in 20 minutes of play. His teammates hit 60%!