Load & Durability
Who is likely to miss games to injury next season, where on the body it tends to happen, and which of the usual suspects (heavy minutes, back-to-backs, travel, miles run, coaches, college workload, body size) actually predict it. Every input had a pass/fail bar written down before it was tested.
What the official injury reports changed
Version 1 had to guess: any run of three or more missed games counted as an injury. The league's own injury reports, filed before every game from 2021-22 on, say why a player was out. Matched against those reports, the guess was mostly wrong.
Games missed to injury by body area, 2021-22 to December 2025, all players.
Projected injury games,
Expected games missed to injury or illness out of 82, for every player on a 2026-27 roster, offseason moves included. The number comes from his injury games in his most recent season with official reports, the only input that passed its test at the season level. A player who missed all of last season is projected from the reports he was listed on; rookies and players with no NBA injury record get the league average and are tagged. Click a player for his body map, his last injury and his workload.
Game by game: what raises the risk
Season totals hide load. So this looks at each of games a player started healthy and asks whether an injury absence started at his next game. Load passed its test here (). The pattern is total minutes, not spikes.
Chance an injury absence starts at the next game, by fifths of minutes played in the previous week.
The week's minutes compared with his 4-week average, in tenths. The jump players take does not add risk; the top tenth is lower than the middle.
Teams list planned "injury management" absences mostly on second nights, so part of this gap is scheduled rest.
Injuries come back to the same place
After a player returns from an injury absence of three or more games, his next injury within 30 games hits the same area of the time. If injuries landed at random in the league's usual mix, that would be . The red line is that random rate.
Do long playoff runs catch up with players?
Long runs mean more games, harder games and bigger minutes: . The catch in testing it is that the players who go deep are usually the good, healthy ones, and a player who gets hurt in the playoffs logs fewer playoff minutes. So these tests use how far his team went, control for age, minutes and his own injury history, and also compare each player with himself. Sample: .
By how far his team went. Bars are the raw rate; the bracket is its 95% range.
Odds of missing 20+ next season compared with players whose team missed the playoffs. 1.0x = no difference.
Team playoff games over his last three seasons, and the share who missed 20+ games the season after.
Each cell: his team's playoff games that season → games he missed the next season ("all" = the whole season).
Games missed the next season count runs of three or more straight games, any reason (official injury reports only start in 2021-22). Thompson's 2019-20 (ACL) and Durant's 2019-20 (Achilles) were whole seasons lost. Iguodala's 2019-20 was a contract holdout in Memphis, not an injury, which is why games missed for any reason can mislead.
The game asks for more running
Coaches: who rides his starters hardest
Coach load is how many minutes a night his five most-used players averaged, compared with the league that season, over every season he was a head coach since 2010-11. Rest is one- and two-game breaks per rotation player per season. Coach load did not forecast injuries at the season level (test L3), so this describes style.
Teams
Nobody publishes how teams train or rehab players, so these lines measure what the work produces. Injury games are from the official reports for 2024-25. A team's medical record did not repeat from one season to the next (test TM1), so read these as last season's luck as much as anything.
The test record
Each line is a question, the bar it had to clear (written before the test was run), and what happened. The full rules, including the fixes made after a first run and the results before them, are in LOAD_PREREG.txt.
| Test | Question | Bar | Result | Verdict |
|---|
Version 1 tests (games missed for any reason, 2010-11 on)
| Test | Question | Bar | Result | Verdict |
|---|
Key
What this does not do
- Season totals are still hard to forecast. A fuller season model (history, age, position, workload, contact, size, a tree model) did not beat last season's injury games alone. The number on the table is honest about that: it mostly says who was hurt recently.
- Load works at the game level, but it is partly exposure. More minutes means more time on the floor to get hurt. The data cannot separate tired legs from simply playing more.
- 2025-26 is partial. NBA.com's firewall blocks its 2026 report files, so the season is covered through December 21.
- High school is not in it. There is no public record of high-school injuries or minutes. College is shown as workload and games missed, and it did not forecast.
- Teams list what they choose to. "Injury management" on a back-to-back is often planned rest, and some injuries are listed vaguely or late.
- The playoff tests use any games missed before 2021-22. Official injury reports only go back that far, so the 15-season playoff sample counts runs of three or more missed games for regulars, which were about 75% injuries when checked against the reports. A holdout or a suspension still counts.
- It is a forecast for groups, not a diagnosis. A 50% chance of missing 20 games means half of players like him did.