Click a game to compare the two teams.
Spreads shown from the home team's side. A negative number means the home team is giving points.
Every standard line on the board beside a projection built from the player's own game log, the defense he is facing and the score the market expects. The posted line counts as evidence too, so a number only stands out here when a lot of games disagree with it.
Game log. Last season and this one. Recent games count more (half-life ten games), this season counts one and a half times, games for a previous team count half.
Opponent. What that defense has allowed per game against the league, taken halfway and capped at 15%.
Game script. The team's implied points from the spread and total against its own scoring average, plus a small run or pass tilt from the spread.
The line is evidence, and it is a median. Yardage is skewed, so a book's number sits under the player's average. The line is converted to the average it implies, then the projection is pulled toward it as if it were twelve more games. If the model has nothing to add, the answer is exactly 50%.
Probability. Gamma for yards, negative binomial for counts, Poisson for touchdowns and interceptions. Whole-number lines can push, and pushes are taken out of both sides.
Not picks. Thin samples, players with no games this season, and lines far from a player's history (the book pricing a role change) are shown greyed and never logged.
2024 and 2025 actual performance, not a 2026 projection. Roster moves since do not appear here. Click a column to sort; small grey numbers are league rank.
Off EPA is expected points added per play. Every situation has a point value based on down, distance and field position, and this is how much that value changed. It is the honest version of yards per play: three yards on third and eight is a negative play. Above +0.10 is a very good offence. Pass and Rush split it by play type.
Success is the share of plays that gained enough to stay on schedule, roughly 40% of the needed yards on first down, 60% on second, all of it on third.
Those four use neutral script only, meaning plays with the game still competitive, so a team that spent November losing by 24 is not flattered by garbage-time yardage.
3rd down is conversion rate. RZ TD is the share of red zone trips finished with a touchdown and RZ trips is how many they get per game. Those answer different questions: how often you get there, and whether you finish.
Def EPA is EPA allowed per play, so negative is good. 3rd allowed and RZ allowed are the same rates from the defence's side, lower being better. Ranks already account for direction, so 1st always means best.
Plays is offensive plays per game. That is pace, not quality. Fast teams are not better teams, they just create more of everything.
Every game since 2014 with an opening and closing number on record.
What each sportsbook is offering right now, the best number available on each side, and how far apart the market is. Nobody has to be right about anything for this to be worth reading. The best number on each side, and which book has it.
Every metric we track, for one team, ranked against all 32. Pick a second team to put its numbers beside them, which is how you read a coaching change or a divisional matchup.
Fifty team metrics, put through two tests. Does it repeat itself next season, and does it predict next season's points. Most of what gets talked about fails both.
One opinion beside the data. It does not beat the market, and its full record is on the home page.