Omaha 1.8.0
Released 10 September 2026
The same team does not call the same game every week. The simulation was treating each team's tendencies as settled, so it produced a narrower range of games than football actually produces.
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 1.7.0’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 1.7.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.1% | +0.12−0.21 to +0.17 |
| Rushing touchdownsrare event | +17.3% | +0.09+0.23 to +0.84 |
| Receiving yards | +13.7% | −0.04−0.10 to +0.02 |
| Pass attempts | +13.0% | +1.81+1.07 to +2.48 |
| Rushing yards | +12.5% | +0.51+0.37 to +0.63 |
| Passing yards | +11.0% | +1.03+0.46 to +1.59 |
| Completions | +9.7% | +1.66+0.96 to +2.29 |
| Passing touchdownsrare event | +6.9% | +0.35−0.28 to +0.38 |
| Rush attempts | +3.9% | +1.42+1.14 to +1.65 |
| Targets | +1.5% | +0.43+0.32 to +0.54 |
| Receptions | +1.3% | +0.06−0.02 to +0.16 |
| Average across all figures | +10.1% |
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
A coach game-plans. He looks at the opponent, the weather, who is healthy, and calls a different game than he called last Sunday. The simulation had no way to express that: it drew every snap of every simulated game from one fixed set of tendencies per team, so the only week-to-week variation it could produce was the randomness of individual plays. That is real variation, and it is not enough of it.
Measured on 1,139 team-games, the simulation reproduced 84% of the real week-to-week spread in how often a team passes, 83% of the spread in how many snaps it runs, and 76% of the spread in pass attempts. Three numbers missing by about the same amount is one missing idea, not three separate faults.
The simulation now draws a game plan once per simulated game and calls the whole game from it. Because the draw moves how often a team throws, it moves the ball between the quarterback and the backfield, which is why it reaches the running stats as well as the passing ones.
Pass attempts improved by 1.8 percentage points, completions by 1.7, rush attempts by 1.4, passing yards by 1.0. Six of twelve stats improved by a significant margin, one got slightly worse, and the published average went from 9.4% to 10.0% better than a season average.
This is a calibration fix rather than a simple widening. Our intervals were too narrow on the passing stats, which meant real results landed outside them more often than they should have. They now land where they are supposed to.
On the passing stats, the measure of how evenly real outcomes fall across our predicted range improved roughly fourfold.
What this version still gets wrong
- The size of the week-to-week variation we add is deliberately about 60% larger than the variation we measured. It is the only lever we have for a spread that is missing from several places at once, so it is doing work that belongs elsewhere. We would rather say so than present a borrowed number as a modeled one.
- That has a direct consequence for anyone reading the spread itself. Our distribution of how much a team's passing volume might vary is wider than we believe football is. Read the projections; do not read that range as our view.
- Because this variation does not carry from one week to the next in real football, nothing about it is predictable. We can say games vary this much. We cannot say which way this Sunday.
- Receiving yards came down by 0.07 of a percentage point, a small but real cost.