Hindsight is undefeated
25 seasons of matchups, reconstructed down to the individual player-week. 561 losses would have been wins with one lineup change — but only about half of those are calls anybody would actually second-guess.
The closest game in league history
0.05 of a point, across 2,358 decided games — 2,383 played, 25 of them tied.
Of the 2,211 losses where the lineup can be reconstructed, 561 would have flipped on one legal substitution — 25% of the losses we can check. Which sounds damning until you ask what the coach actually knew on Sunday morning, and the pile splits almost evenly in two.
The player on the bench had been the better player all season. These are the starts that are genuinely hard to defend — the coach had the evidence and went the other way.
And these are the other 300: the coach started his better player, by season average, and the week simply went the other way. You do not bench an elite stud for somebody's WR3, and nobody in this league ever has. A stud who lays an egg is not a bad decision, it is Sunday.
Legal is doing real work in all of this. A substitution counts only if the eight that result still fill that era's slots, matched against the league's own lineup — so you cannot bench a defence for a receiver. The CBS era changed shape four times between 2002 and 2020 and each season is checked against its own. The split between the two piles is the season average each player was carrying that year, which is the closest thing the archive has to what a coach knew at the time.
Every decided game in the archive, by final margin.
With perfect hindsight — the best legal eight the roster could have fielded that week — 833 of those losses turn into wins, and the average team-week leaves 14.3 points on its own bench. These are the five worst.
| season | team | result | scored | best legal | left behind |
|---|---|---|---|---|---|
| 2020 w10 | Gamblin Gars | lost | 25.7 | 106.5 | +80.8 |
| 2025 w17 | Cunning Stunts | lost | 65.5 | 132.45 | +66.95 |
| 2013 w12 | Blackbeards | lost | 30 | 96 | +66 |
| 2003 w6 | Knights Who Say Ni | lost | 38 | 103 | +65 |
| 2022 w17 | The Nasty Boyz | lost | 60.5 | 124.25 | +63.75 |
One week per position, because a single list is six quarterbacks and tells you only that quarterbacks score more. Scored under the rules in force that season.
One-swap losses per coach against games coached, for everyone with at least 100 games. A rate, not a verdict — you only get one of these by being close enough to lose narrowly.
| coach | one-swap losses | games | rate |
|---|---|---|---|
| Ryan Ayers | 28 | 170 | 16.5% |
| Jacob Weissburg | 36 | 221 | 16.3% |
| Steve Baird | 16 | 102 | 15.7% |
| Joshua Liebman | 29 | 197 | 14.7% |
| Brian Pugh | 33 | 246 | 13.4% |
| Mike Pugh | 41 | 317 | 12.9% |
| Dan Weissburg | 42 | 328 | 12.8% |
| Kevin Herndon | 40 | 315 | 12.7% |