Omaha 1.2.0
Released 4 September 2026
The offensive line now counts. Overall accuracy is unchanged; late-game accuracy improved by half a point.
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 1.1.0’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 1.1.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.1% | −0.26−1.11 to −0.08 |
| Rushing touchdownsrare event | +17.5% | −0.10−1.39 to +1.04 |
| Receiving yards | +12.8% | +0.00−0.10 to +0.09 |
| Rushing yards | +10.9% | +0.16−0.12 to +0.40 |
| Pass attempts | +9.5% | −0.07−0.29 to +0.16 |
| Passing yards | +9.4% | −0.23−0.80 to +0.34 |
| Completions | +7.2% | +0.09−0.24 to +0.41 |
| Passing touchdownsrare event | +6.6% | −0.79−3.10 to +1.27 |
| Receptions | +1.3% | +0.02−0.09 to +0.08 |
| Rush attempts | +1.2% | +0.03−0.28 to +0.27 |
| Targets | −0.4% | +0.03−0.04 to +0.11 |
| Average across all figures | +8.7% |
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
Until now the simulation judged how good an offense was from its quarterback, backs and receivers only. The five linemen, half the players on the field, and the ones who decide whether a run has anywhere to go, counted for nothing. They now contribute like everyone else.
Measured on plays the model had never seen, offensive linemen predict how a run will go slightly BETTER than the skill players do. On its own this changed almost nothing: rushing yards improved 0.2 points and the overall figure was flat.
Adding five more players to the calculation diluted the quarterback's share of it, from 47% to 38%. Restoring his weight turned out to help RUNNING plays rather than passing ones. His individual rating is the single best read we have on how good an offense is overall, including on plays he hands the ball off.
Late-game accuracy across all eleven figures improved from 16.5% to 17.0% better than a player's own season average. Validated by fitting the weight on the first half of the season and testing it on the second.
Overall accuracy moved from 8.9% to 8.8%, which is no change. This release is worth having for what it makes possible rather than what it delivered.
What this version still gets wrong
- This release did not improve our numbers overall. We are keeping it because judging an offense without its linemen is indefensible on its own terms, and because it is the foundation for work we intend to do. If that work does not pay off, this is a candidate to remove rather than build on.
- We use each team's five most-used available linemen rather than the exact five on the field for a given play. The line is close to constant, so this costs little, but it is an approximation.
- Targets, at −0.4%, remains the one figure slightly worse than a player's own season average.
- We corrected a reporting error found while investigating this release: passing touchdowns were being given a secondary score meant for rare events, which they are not, half of them are non-zero and a fifth are two or more. The main figure was never affected; the secondary one is no longer shown.