Who Gets the Minutes? Inside Our NBA Injury Opportunity Model

Every injury in the NBA frees up minutes. Thirty of them, sometimes more, handed to someone else on the roster every night the player sits. In fantasy basketball, those minutes are the most valuable thing on the waiver wire, and they go to whoever finds them first.

The trouble is that the obvious answer is usually wrong. The backup listed behind him on the depth chart is the first name everyone reaches for, and coaches rarely hand the minutes over that neatly. They spread them across the rotation, move a starter to a new position, or give a run to someone nobody was watching.

So we built a model to answer the question a manager actually has to answer: when he sits, who plays more, and by how much?

One question for every player on the report

For every rotation player on the NBA's official injury report, the model looks at each of his teammates and estimates how many more minutes that teammate will play if he sits, and what those minutes are worth in a nine-category league. It does this for every game, from the evening before until tip-off. A player who barely plays frees almost nothing, and a rookie who has yet to play has no minutes to hand out, so for them it simply says so.

When it finds a teammate it expects to play at least six more minutes than usual, it names him. When it does not, it says so. An injury that frees no real opportunity is information too: it tells you not to burn a waiver claim chasing it.

Learned from four seasons of real rotations

The model learned from every regular-season game of four full NBA seasons, 2021-22 through 2024-25: more than 19,000 games missed by rotation players, and over 200,000 cases of a teammate playing without them. In each one, it saw what the rotation did in reality, not what a depth chart said it would do.

Every signal that went into it had to earn its place. Each one was tested on seasons the model had not learned from, and kept only if it made the predictions better time after time. Several ideas that sound convincing on paper did not survive that test and were left out. The full recipe stays with us. What we can say is that it reads only what is known before tip-off. Nothing from the future, nothing a manager could not have known at the time.

Tested on a season it never saw

Any model looks good on the games it learned from. The only honest test is games it has never seen, so we held back the entire 2025-26 season: 5,773 games missed by rotation players, and more than 59,000 cases of a teammate playing without them. The model never saw one of them while it was built.

Its first pick for each absence played more minutes than he usually did 87% of the time. Seven times in ten, he gained five minutes or more. More than half the time, he gained ten or more. On average, its first pick played 11 more minutes than usual.

Take its top two picks, and at least one of them played more than usual 95% of the time, and at least one gained five or more minutes 84% of the time.

For comparison, a teammate picked at random played more than usual 61% of the time and gained three minutes on average. Taking the teammate who usually plays the most minutes at the injured player's position did no better than that: more minutes 63% of the time, two and a half on average. Even the sharpest shortcut we tested, betting on whoever stepped up the last time the same player sat, fell well short at 80%.

When it says ten minutes, it means ten minutes

Naming the right player is half of it. Knowing how big the opportunity is matters just as much: a ten-minute bump changes a roster, a three-minute bump does not. We checked every first pick against what actually happened:

When it expected a gain under five minutes, its pick averaged three and a half more. Between five and eight, he averaged nearly seven. Between eight and twelve, he averaged ten, and played more than usual nine times in ten. When it expected more than twelve, he averaged nearly eighteen more minutes, played more than usual 96% of the time, and gained five or more minutes 91% of the time.

The size of the call is the size of the opportunity. A big number is a pickup; a small one is a player to watch.

Not just tonight

An injury rarely lasts one game. That is why this model works hand in hand with our injury return model, which gives each injured player's chance of playing in each of his team's next five games. Together they turn a single night's opportunity into the opportunity over the coming week: the minutes a teammate gains in every game the player is likely to miss.

On the 2025-26 season, those combined estimates of a teammate's extra minutes over the next five games were off by about 14 minutes on average, against 19 for the common shortcut of assuming the injured player misses all five. That is the difference between a pickup for one night and a pickup for the week.

Ready from opening night

The first weeks of a season are the hardest time to read a rotation: new signings, trades, players who have never shared the floor. The model handles those too, drawing on what each player did the season before. Because there is less to go on, it holds those calls to a higher bar and names a teammate only when it expects a much bigger gain. As the games come in, its picks sharpen with them.

Live, all season long

The model runs by itself throughout the season. Every time the NBA publishes a new official injury report, the picks are recalculated within minutes, using each team's current roster. A player downgraded to Out in the afternoon has his teammates' opportunities ready well before tip-off. Every prediction is kept with the moment it was made, so its record can be measured against real outcomes as the season goes on.

What no model can see

No model sits in on a coaching meeting. A blowout that empties the bench, foul trouble, or a decision to give a young player a look can move minutes either way on any given night. That is why the model names a teammate only when the opportunity is real, and tells you how big it is, rather than promising a number it cannot guarantee.

Where to find it

The picks sit on our NBA Injury Report, under "If he sits" for every player on the list, beside his chance of playing each of his team's next five games. Check it before you make your next waiver claim.