When Will He Play Again? Inside Our NBA Injury Return Model

"Out" tells you almost nothing. A player listed out tonight might be back on Wednesday or gone until March, and the injury report reads the same either way: a body part, a word like "soreness", and no date. For a fantasy manager, that gap is where weeks are won and lost. Drop him too early and someone else gets the comeback. Hold him too long and the roster spot costs you every night he sits.

So we built a model to answer the question that actually matters: will he play, and when?

One question, five games ahead

For every player in an injury absence, the model gives his chance of playing in each of his team's next five games. It does not hand you a single date that is either right or wrong. It gives you five probabilities that show how sure it is: 3% for the next game, 40% for the third, 85% for the fifth. That shape is the information. A steep climb means a short absence with a real return in sight. A flat line near zero means stash him or let him go.

Probabilities matter because injury returns are genuinely uncertain. A forecast that pretends otherwise, like "back on the 14th" with nothing attached, is a guess in a confident voice. Ours tells you how much to trust every call.

Built on every official report

The model learned from four full NBA seasons, 2021-22 through 2024-25: every official injury report the league published, more than 5,000 injury absences and over 66,000 player-nights. Each absence is followed all the way to the night the player actually returned.

Every one of those nights is a lesson in what the report said, what it left unsaid, and what happened next. Across thousands of absences, patterns appear that no single report shows: which situations tend to mean days, and which tend to mean weeks. The full recipe stays with us. What we can say is that it only ever reads what was known on the night of the decision. 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 data it learned from. The only honest test is games it has never seen. So we held back the entire 2025-26 season: 1,581 injury absences and 18,362 player-nights. The model never saw a single one of them while it was built. Then we asked it, night by night, the same question you would.

For the very next game, it was right 88% of the time.

On nearly two nights out of three it made a confident call, a chance under 10% or over 90%. Those calls were right 97.6% of the time.

It keeps its edge further out, too. Five games ahead, where even teams rarely commit to a timeline, it was still right three times out of four.

And on the very first night of an absence, before a single game had been missed, it placed the return correctly as this week, this month or longer 73% of the time.

When it says 20%, it means 20%

Making the right call is one thing. Getting the number right is harder, and it is what makes a probability worth using. We checked every next-game chance the model gave during 2025-26 against what actually happened:

When it said 10% or less, 2% of those players played. Between 10% and 30%, it averaged 19% and 20% played. Between 30% and 50%, it averaged 41% and 44% played. Between 50% and 70%, it averaged 61% and 63% played. Between 70% and 90%, it averaged 83% and 82% played. Above 90%, it averaged 97% and 96% played.

That is what calibrated means. A 20% really is a 20%. You can set a lineup by it, weigh a stash against a waiver pickup, or decide how long to hold a player, with the real odds in front of you.

Live, all season long

The model runs by itself throughout the season. Every time the NBA publishes a new official injury report, the chances are recalculated within minutes, so a player upgraded from Out to Questionable in the afternoon has new numbers well before tip-off. Every prediction is stored 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 sees inside a locker room. A surprise setback, a team quietly resting a healthy player, or a call made an hour before tip-off can still move a return either way. That is exactly why the model speaks in probabilities rather than promises: it tells you when it is sure, and when it is not.

Where to find it

The model's chances sit on our NBA Injury Report, beside the return dates ESPN and CBS Sports publish, so you can see the sources' estimates and our odds side by side. Check tonight's injured players there before you set your lineup.