Omaha 2.5.0
Released 10 October 2026
The simulation now tells a quarterback scramble from a designed quarterback run, and every running play fields the players its formation names. Rush attempts, rushing yards and rushing touchdowns all improve, and the average rises from +10.0% to +10.8%, the highest it has been.
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 2.4.0’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 2.4.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.2% | +0.03−0.46 to +0.53 |
| Rushing touchdownsrare event | +18.1% | +1.40+1.24 to +4.15 |
| Receiving yards | +14.8% | +0.39+0.17 to +0.67 |
| Pass attempts | +13.0% | +0.00−0.13 to +0.14 |
| Rushing yards | +12.7% | +1.80+1.07 to +2.57 |
| Passing yards | +11.1% | +0.03−0.11 to +0.16 |
| Completions | +9.7% | −0.04−0.20 to +0.11 |
| Passing touchdownsrare event | +7.1% | −0.11−0.97 to −0.05 |
| Rush attempts | +4.1% | +3.21+2.12 to +4.51 |
| Targets | +3.9% | +0.73+0.34 to +1.22 |
| Receptions | +3.7% | +0.45+0.07 to +0.86 |
| Average across all figures | +10.8% |
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
A formation called "11" means one running back, one tight end and three receivers. Since 2.0.0 the simulation worked out how many of each to field by rounding each position on its own, and a back who was only partly expected on the field could round down to nobody. Those runs went to whoever was there instead, often the quarterback or a receiver. Each formation now fields exactly the players its name says.
In one matchup we checked, 7% of simulated runs had no running back on the field. Running backs took 76% of carries where the engine before 2.0.0 gave them 84%, and quarterbacks 17% where it gave 12%. This was the unexplained rushing loss recorded in 2.0.0. Fixing it alone recovered most of that loss.
A scramble is a pass play that broke down, and it is recorded as a run. The simulation treated it as a handoff and drew its result from ordinary running plays. It now decides whether a run is a scramble, from the down and distance, the score, the clock and the quarterback, and draws its yardage from real scrambles.
Quarterbacks scramble on 45% of runs on third and long in real games and 45% in the simulation. Trailing by two scores in the fourth quarter it is 15% real and 17% simulated. Measured on 2025 snaps the simulation had not been trained on.
When a team needs one yard, the quarterback very often keeps it himself. The simulation decided who carried the ball the same way at one yard to go as at ten, so it rarely called the sneak. It now decides separately whether the quarterback keeps a designed run, and when he does not, the handoff goes to someone else.
At one yard to go, quarterbacks carry 18.4% of designed runs in real games. The simulation gave them 7.8% and now gives 18.0%, gaining 1.98 yards a carry against a real 2.00.
The three changes above all affect who carries the ball, so the rushing figures move the most. Receiving figures move up slightly as well, because the same fix puts the right number of receivers and tight ends on the field.
Rush attempts +3.2 points, rushing yards +1.8, rushing touchdowns +1.4 (and +2.4 on the touchdown probability score). Targets +0.7, receptions +0.5, receiving yards +0.4. Measured across 2024 against the same baseline as 2.4.0.
What this version still gets wrong
- Passing touchdowns move down 0.1 points across all games, and 0.5 points on a game-by-game comparison. Both are small, and we do not have an explanation for them yet.
- Designed quarterback runs still gain too much: 3.7 yards a carry against a real 3.1. Away from one yard to go they are also called a little too often, partly because the league changed: quarterbacks took 6.7% of designed runs in 2024 and 5.1% in 2025.
- Scrambles run a little high overall, about 10% to 15% more often than real.