The line on each card fills with how often the loser wins the replay: the fuller, the more luck decided it. Outlined: the winner was outplayed.

Week 3, 2026

Week 2, 2026

Week 1, 2026

The adjusted standings, 2026

Order by Group by

AFC

The same seven teams make the playoffs if the season ended today.

NFC

With the luck taken out, the Cowboys and the Buccaneers would make the playoffs if the season ended today instead of the Giants and the Panthers.

Adjusted wins: every kick, fumble bounce, drop, conversion and referee call re-rolled at your settings, everything else as played. Seed: playoff seed in the adjusted standings; 5 in only in the adjusted standings (luck kept them out), out is in by record only thanks to luck. Adjusted: rank in the conference by adjusted wins, ▲ places up because luck cost them, ▼ down because luck helped. Now: rank by record.

Luckiest so far

All 32 teams, after Week 3 · Power rankings: who is best with the luck taken out

The biggest heists of 2026 so far

- Broncos 30, Rams 26 · Week 3 · the Broncos win 7% of replays

- Browns 21, Panthers 18 · Week 3 · the Browns win 9% of replays

- Giants 28, Cowboys 20 · Week 1 · the Giants win 15% of replays

- Colts 19, Texans 17 · Week 3 · the Colts win 25% of replays

- Saints 24, Ravens 17 · Week 2 · the Saints win 26% of replays

How it works

Every play of every game goes back into the pot. Each team’s own plays from that game are drawn at random, by down and distance, into new drives and a new game, 4,000 times. Kicks, fumble bounces and turnovers happen at the rates you would expect, not the ones that happened. How often a team wins those replays is what its performance was worth that day.

The difference between that and the result is luck. A loss in a game a team wins 9 times in 10 scores −90 on our luck index. Kicks, fumbles, drops, conversions and referee calls are valued one by one, against each kicker’s and receiver’s own record, so you can see where it came from.

The replay predicts real rematches about as well as the betting line does. It does not predict who covers next week: the market already knows. Methodology and validation