Omaha 1.6.0
Released 4 September 2026
Before kickoff, the simulation was too sure of itself. It now leans harder on what a player has actually done this season, and every figure improved or held.
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 1.5.0’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 1.5.0 |
|---|---|---|
| Receiving touchdownsrare event | +20.5% | +0.00+0.00 to +0.00 |
| Rushing touchdownsrare event | +17.3% | +0.00+0.00 to +0.00 |
| Receiving yards | +13.5% | +0.09−0.06 to +0.24 |
| Rushing yards | +11.7% | +0.59+0.34 to +0.86 |
| Pass attempts | +11.2% | +0.37−0.08 to +0.79 |
| Passing yards | +10.0% | +0.23−0.18 to +0.59 |
| Completions | +8.2% | +0.14−0.34 to +0.57 |
| Passing touchdownsrare event | +6.6% | +0.00+0.00 to +0.00 |
| Rush attempts | +2.1% | +0.79+0.46 to +1.14 |
| Receptions | +0.9% | −0.06−0.23 to +0.11 |
| Targets | +0.6% | +0.38+0.17 to +0.60 |
| Average across all figures | +9.3% |
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
With a game underway the simulation has the score, the clock and the situation. Before kickoff it has none of them, and its estimate of a player's workload is built from several uncertain pieces multiplied together. We already corrected for this by leaning on the player's own season average, but not nearly enough: measured now, the simulation was departing from that average about twice as far as the evidence justified.
Rush attempts improved 1.2 points on held-out weeks, rushing yards 0.9, targets 0.5. Nothing got worse.
The correction is not the same everywhere, and the reason is simple. A quarterback takes essentially every pass attempt, so there is no question of WHICH player, only how many. Every other figure requires picking a player out of several, and that is where the simulation was overconfident. So passing figures are now corrected less, and everything else more.
Two settings rather than one per figure. Tuning each figure separately scored worse on weeks we held back, which is what fitting to noise looks like.
Touchdowns are left alone, as they have been since we introduced this correction. They happen rarely enough that a player's average over part of a season is itself too noisy to correct toward.
All three touchdown figures moved by exactly 0.00.
What this version still gets wrong
- This is not the simulation getting better at football. It is the simulation being more honest about what it does not know before a game starts, by leaning harder on history. In modeling terms that is a step backwards taken because it is measurably more accurate, and the correction shrinking over time is how you will know the underlying model has improved.
- The gain is concentrated later in the season. In the first few weeks a player's season average is itself built on very little, so leaning on it harder helps less, two figures were slightly worse across weeks 1-9.
- Receptions, at +0.9%, and targets, at +0.6%, remain the two figures closest to simply looking up a player's season average.
- Our pre-game figures for counts, targets, receptions, carries, are still the weakest thing we publish. During a game they are strong; before one, they are close to worthless, and this correction narrows that gap rather than closing it.