Predicting Playoff Injuries
I hope I'm wrong!
Nobody likes to think about injuries, and I certainly hope this article doesn’t jinx any player. But, health can have as much of an impact on playoff success as anything else, so I think it’s worth exploring.
Take the 2018-2019 Warriors. The “Death Lineup” finished first in the Western Conference and were favorites to win it all, but then the injuries started. DeMarcus Cousins tore his quad in Round 1, Kevin Durant strained his calf in Round 2, and Andre Iguodala strained his calf in Round 3. The Warriors made the Finals anyway, but Durant (recently returned) ruptured his achilles in Game 5, then Klay Thompson tore his ACL in Game 6, and the Warriors fell to the underdog Toronto Raptors.
On the other side, most champions remain fairly healthy through the playoffs. The 2023 Denver Nuggets won the title partly because they were (finally) healthy. Following years of Jamal Murray knee and ankle injuries, he and the rest of the team made it through the playoffs without missing a game. Of course they still had to be very, very good to take down Anthony Edwards, Devin Booker, LeBron James, and Playoff Jimmy, but being healthy didn’t hurt.1
The data journalism site, The Pudding, has actually tracked how much each NBA champion was helped by injuries in what they call the The CRUTCH2 Rankings. The Nuggets championship team ranks 3rd in CRUTCH, only behind the 2013 Heat and the 2015 Warriors.
Given the importance of injuries, I wondered, using publicly available infomation3 is it possible to predict who is more or less likely to get injured in the playoffs? And if so, who is most at risk this year?
I won’t bury the lead. Here is my ranking of playoff injury probabilities:
The top of this list generally makes sense, with a mix of older players (Steph Curry, Al Horford, Mike Conley), players with lengthy injury histories (Ben Simmons, Ja Morant), vets who have carried a big load for their teams this year (Norman Powell, Jamal Murray), and smaller, scrappy guards (Gary Payton II, Fred VanVleet).
Here’s what the injury risk looks like at the team level, weighting players by the number of minutes per game they played during the regular season:
I will readily admit that the regression used to make these predictions is not a super close fit4, which just goes to show that predicting injuries is hard! For a few reasons:
There are relatively few of them
A lot of injuries in fast-paced, physical games are just a matter of fluky landings or crashes that simply can’t be predicted.
In the final push for a ring, players often opt to play through injuries, and thus do not show up in the data as actually being injured. Jamal Murray is officially credited with 0 career playoff injuries, but he certainly hasn’t been 100% in all those games.
Nevertheless, the metrics described above (age, injury history, minutes played, position) turn out to be at least somewhat correlated with playoff health. Put all of them (and more5) into a regression and out pop the predictions for this year.
Sadly, Substack cannot embed Flourish charts, so click here to see the details Here’s a sneak peak:
I’ll check back in at the end of the playoffs with how accurate this is, and again, I hope not at all accurate and somehow this is the year that every player makes it through the playoffs in good health6 🤞
Literally.
Championship Revisions Utilizing Top Competitor Health.
Obviously the players, trainers, and coaching staffs have a lot more information and would be able to make much more accurate predictions, as this book proposes.
For those curious about the quality of the model, the McFadden pseudo R2 is 0.26 (not bad). On the training data, the model assigns <0.25 injury probability 99% of the time when a player did not actually get hurt in the playoffs and >.25 injury probability 63% of the time when a player did actually get hurt in the playoffs. Age, days missed due to injury during the regular season, and the number of days before the playoffs a player returned from injury were statistically significant at the 0.05 level.
The logistic regression includes a player’s age, position, number of teams played for during the regular season, basic counting stats (games, minutes played, offensive rebounds, and points) during the regular season and in a career overall, whether they were an All-Star, prior injuries and days missed during the regular season, in previous playoffs, and in a career overall, and the number of days before the playoffs a player returned from their last regular season injury.
I tried other stats as well but none were particularly predictive. In the future I may add other factors such as a player’s schedule (back-to-backs, miles traveled), height/weight and changes year over year, etc.
Especially Jamal Murray.