Omaha 1.3.0
Released 4 September 2026
The simulation now runs a two-minute drill at two-minute-drill speed, kneels out a win, and keeps track of timeouts. Ten of the eleven figures we publish improved.
How to read these numbers
Every figure below compares the simulation against a player’s own season average so far, scaled to how much of the game is left, roughly what a reader could work out without us. Zero means we add nothing over that. Positive means we beat it; negative means we are worse than it.
Measured across the 2024 season, weeks 1–18, rebuilding the model each week using only what was known at the time. 400 simulations per situation.
The change column is simply this version’s accuracy minus 1.2.0’s, each measured across everything that version covers, so the two columns you can see always account for it. We also measure the change on shared situations alone, which is a more sensitive test; where the two disagree we publish this one.
Accuracy by stat
| Stat | vs. season average | Change from 1.2.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.5% | +0.37−0.16 to +0.26 |
| Rushing touchdownsrare event | +17.2% | −0.30+0.44 to +1.35 |
| Receiving yards | +12.9% | +0.05+0.04 to +0.13 |
| Rushing yards | +11.1% | +0.21+0.10 to +0.29 |
| Pass attempts | +10.8% | +1.28+0.98 to +1.57 |
| Passing yards | +9.8% | +0.44+0.22 to +0.69 |
| Completions | +8.1% | +0.86+0.60 to +1.14 |
| Passing touchdownsrare event | +6.7% | +0.12−1.86 to −0.53 |
| Rush attempts | +1.5% | +0.21+0.04 to +0.36 |
| Receptions | +1.4% | +0.12+0.07 to +0.22 |
| Targets | −0.2% | +0.18+0.10 to +0.27 |
| Average across all figures | +9.1% |
The average is an unweighted mean across the figures above. It is a summary of how we are doing, not a single score for the model: yards and counts are not measured in the same units, so combining them any more cleverly than this would just let passing yards decide the answer.
Each change carries the range the measurement can actually support. A change whose range crosses zero is shown in neutral rather than as a gain: we cannot tell it apart from no change, and we are not claiming it as one.
What changed
Plays take far less time when a team is hurrying, about ten seconds inside two minutes against thirty earlier, and the simulation did not know that. It was fitting eight plays into the last two minutes where real games fit fourteen. The total number of plays in a game was right all along, which is why this went unnoticed: they were simply in the wrong part of the game.
Simulated plays in the last two minutes of a half went from 8.1 to 12.8 against a real 14.5, and every other stretch of the game moved closer too.
Real teams take a knee. The simulation could not, kneel-downs are recorded separately from ordinary plays and were absent from everything it learns from. That did not matter while it was too slow to fit many late plays; once the clock was right, it started running fourteen real plays where a team would have knelt three times and gone home, inventing carries and attempts that never happen.
Fourth-quarter rush attempts were being over-projected by 17% and pass attempts by 21%; with kneel-downs modeled, rush attempts are now within 0.3%.
A team trailing by a score inside five minutes throws on 92% of snaps with no timeouts left and 68% with three. The simulation now tracks timeouts for both sides, spends them at the rate real teams do, and returns them at half-time.
Pass attempt accuracy improved 1.3 points, completions 0.9, passing yards 0.4.
What this version still gets wrong
- Rushing touchdowns fell 0.3 points, the only figure to get worse.
- Timeouts are spent at a league-average rate rather than for a reason. A real team calls one to save the clock on a particular play; ours calls one with a probability that depends on the score and the clock. The distribution of timeouts left at the two-minute warning is close to right, which is what the play calling reads.
- We still project slightly too many pass attempts late, about 12% in the fourth quarter. The play mix and the play count are both correct, so this is something else, and it is not yet understood.
- Targets, at −0.2%, remains the one figure not better than a player's own season average, though it improved again.
- Our rush attempts exclude kneel-downs, which official statistics count. For a quarterback closing out a win this makes our figure lower than one you would look up elsewhere. We intend to align the definition.