Omaha 1.7.0
Released 7 September 2026
When a game stops being competitive, real teams start resting their starters. The simulation was playing its best players to the final whistle.
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 1.6.0’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 1.6.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.0% | −0.37−0.10 to +0.14 |
| Rushing touchdownsrare event | +17.2% | +0.14+0.05 to +0.43 |
| Receiving yards | +13.7% | +0.19+0.16 to +0.32 |
| Rushing yards | +12.0% | +0.26+0.20 to +0.47 |
| Pass attempts | +11.2% | −0.02−0.07 to +0.03 |
| Passing yards | +10.0% | +0.03−0.01 to +0.08 |
| Completions | +8.1% | −0.02−0.08 to +0.06 |
| Passing touchdownsrare event | +6.5% | +0.03−0.03 to +0.24 |
| Rush attempts | +2.4% | +0.38+0.30 to +0.65 |
| Receptions | +1.2% | +0.30+0.26 to +0.48 |
| Targets | +1.1% | +0.44+0.42 to +0.68 |
| Average across all figures | +9.4% |
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 team up three scores late takes its starting running back off the field. It has a lead to protect, a season to protect him for, and nothing left to prove. Real teams do this so consistently that a starting back's share of his team's carries falls from about 63% in a one-score game to about 21% once the margin reaches three scores. The simulation had no way to know: it chooses who is on the field from personnel and play type, and neither of those knows the score or the clock. So it kept feeding its starters, and gave almost nothing to the backups who were actually taking the carries.
Across a decided fourth quarter the simulation was giving starting backs 62% more carries than they took, and backups 12% of theirs. Its response to the game state was 1.4 percentage points where the real one is 63.7.
This matters well beyond the minutes it describes. Every figure we publish is a projection for the REST of the game, so a projection made at kickoff already contains whatever garbage time the game later reaches. Getting the end of a blowout wrong quietly moved every earlier number in the same game.
18.1% of all remaining carries, and 17.2% of remaining targets, occur in a game that has already been decided.
The correction is applied to every skill position, not just running backs, and it is not symmetric: a team that is LOSING badly keeps its starters on, while a team winning badly rests them. Being behind is a reason to keep trying; being ahead is a reason to stop.
Trailing by three scores, a starting back still takes 38% of his group's snaps; leading by the same margin, 28%.
What this version still gets wrong
- This corrects who is on the field, not how hard anyone plays. A team resting starters is also running a different offense, and we do not model that.
- It is fitted from how far a game has gotten out of reach, in five margin bands. A fourteen-point lead and a nineteen-point lead sit either side of a line, and real coaches do not.
- The correction can only spread carries across the players a team actually used. It cannot field a fourth-string back who took no snaps all season, which is sometimes exactly who ends up carrying the ball in the last three minutes.
- Counts before kickoff, targets, receptions, carries, remain the weakest figures we publish. This improves them and does not fix them.
- Receiving touchdowns read 0.37 points lower than in 1.6.0 above, and we do not believe that is a real decline. The more sensitive test, the same games, the same players, the two versions compared directly, puts the change at +0.02 with a range that spans zero, meaning it cannot be told apart from no change at all. The two disagree because they cover slightly different sets of situations. We publish the larger, less flattering number because it is the difference between the two figures a reader can actually see on this page.