Omaha 1.12.0
Released 15 September 2026
Fourth and one and fourth and twelve were being treated as the same situation. So were first and goal from the one and first and goal from the five.
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 1.11.0’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 1.11.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.7% | +0.38−0.49 to +0.05 |
| Rushing touchdownsrare event | +18.1% | +0.05−0.08 to +0.65 |
| Receiving yards | +13.8% | +0.05+0.01 to +0.08 |
| Pass attempts | +12.9% | −0.04−0.15 to +0.08 |
| Rushing yards | +12.6% | +0.05−0.04 to +0.07 |
| Passing yards | +11.2% | +0.15+0.03 to +0.28 |
| Completions | +9.8% | +0.05−0.07 to +0.20 |
| Passing touchdownsrare event | +7.1% | +0.15−4.02 to −1.94 |
| Rush attempts | +4.0% | +0.06−0.00 to +0.14 |
| Targets | +1.5% | +0.01−0.06 to +0.06 |
| Receptions | +1.4% | −0.05−0.10 to +0.04 |
| Average across all figures | +10.3% |
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
The simulation picks what happens on a play by looking up real plays from a similar situation. But its idea of a situation knew the down and not the distance, except on third down. Every fourth down shared one pool, so fourth and one was answered with plays that included fourth and twelve. Near the goal line the same thing happened with field position: first and goal from the one drew on plays from anywhere inside the five.
Those pooled groups are not alike. Fourth and one to three gains 3.85 yards; fourth and four to seven gains 5.43. From the one yard line the simulation was expecting 1.05 yards where real plays gain 0.24.
Splitting situations finely only helps if there are enough real plays to fill each one. So instead of widening the situation when a group runs thin, the simulation now reaches further back through older seasons to find more plays from the same situation. Short-yardage football has barely changed in eight years, so those older plays are still the right answer.
Fourth and one from deep in a team's own half had 82 matching plays in two seasons and has 217 across eight. Yards gained on short-yardage plays moved by 0.21 across 2016 to 2023, with no trend.
Drawing from the right plays moves the numbers those plays feed, so the gains are spread thinly across many statistics rather than concentrated in one.
Passing yards and receiving yards both improved by a significant margin, twelve of the fourteen figures we track moved up, and the published average went from 10.2% to 10.3% better than a season average.
Our receiving touchdown projections got slightly better on one measure and slightly worse on the other. We are shipping it anyway because everything else moved up, but we would rather say so than report only the half that suits us.
Receiving touchdown accuracy improved 0.38 points while the measure built for rare events fell 0.35.
What this version still gets wrong
- The simulation still reaches the one yard line about 41% as often as real teams do. This fixes which plays it draws once it gets there, not how often it gets there.
- Field position near the goal line is still grouped as one zone covering the one through the five, and real scoring chances across that range run from 56% down to 32%.
- With one yard to go the simulation still throws on about half of snaps where real teams throw on a quarter. The fix for that is built and costs us elsewhere, so it is off.
- Reaching back through older seasons assumes the game has not changed much. That holds for short yardage over the window we use, and we have not checked every situation.