Published evidence · 2026
Track record
Every call we published, scored after the game. Nothing is edited after kickoff; corrections are dated notes next to the original.
Every call we published, scored after the game. Nothing edited after kickoff.
Scored Sun Nov 15 · 4:33 PM ET · artifact verified
Season record
Season
87 of 139 · 63%
Every published call, coin flips included. A lean counts as a call.
coin flips counted
Against the market
Vs the market
−7 games
Same 138 games, 1 tie out: the closing favourite went 94 of 138. Market line, research only.
Market line, research only.
Brier score
Brier
0.231
Lower is better. A coin flip scores 0.250; a perfect forecaster 0.
coin flip is 0.250
Not counted
0
Nothing to leave out yet.
nothing to leave out
This view: week 10 and the calls still to be scored. Every week: week pages · CSV
Calibration, what we said vs what happened
n = 139Grey tick: the confidence we published. Bar: how often that side won. A bar past the tick means we were underconfident; short of it, overconfident. Buckets under 3 games are shown in grey and prove nothing yet.
Calibration, said vs happened
Tick: what we said. Bar: how often that side won. Grey: under 3 games.
Season, week by week
Each tile is one week: hits and misses on top, and the tile fills once every game that week is scored. The season Brier and the calibration table update the same night.
Us against the alternatives, same 138 games, 1 tie left out
Skill: how much better than a coin flip the Brier is, in percent. Closing line: the market's closing favourite. Closing line from the market, shown for research. Not our number and not advice. The closing line is the hardest benchmark there is; we report where we stand against it.
Us vs alternatives, 138 games, 1 tie left out
Bar: skill vs a coin flip. Closing line: the market's closing favourite. Closing line from the market, shown for research. Not our number and not advice.
The margin range, tested
The margin range, tested · 133 of 139 inside
133 of 139 inside · 96%We say 9 in 10 finals land inside the bar. Each row is a Week 10 final placed on its own range: inside is green, outside is red, the tick is our median.
Scale: margin from the side we called, −25 to +35.
Grey: our range. Tick: our median. Dot: the final, green inside, red outside. Scale −25 to +35.
Player projections, scored after Week 9
3701 projections published · 2936 scoredBiggest misses
Best calls
Not published yet.
MAE: mean absolute error in fantasy points, half PPR. Inside: the share of players whose actual landed between the floor and ceiling we published, our 20th and 80th percentiles, so about 60 in 100 should; within 10 of that is green, outside it amber. Rank correlation: each week's Spearman between our order and the finish, averaged over the weeks; 1 is perfect. Every number is scored against the official box score once it is published.
Player projections, Week 9
MAE in fantasy points, half PPR. Inside: share of players whose actual landed in our floor to ceiling, the 20th to 80th percentile; about 60 in 100 should.
14 of 150 calls on this page · export CSV · the published artifact
Margin error: final margin minus our median, from the side we called. Brier per game: (confidence − outcome)². ✎ is a dated correction; the call itself is never edited.
14 of 150 calls on this page · export CSV
Err: final margin minus our median. Brier: (confidence − outcome)². Calls are never edited after kickoff.
How this page works
A published call is never edited or deleted after kickoff. A correction is a dated note next to the original. Every row keeps its Hit, Miss, Pending or Not counted so you can check us; the Brier score checks both the outcome and the confidence we stated. A near coin flip still counts: if the side we leaned toward wins, it is a Hit; if not, a Miss. A tie is a Miss, because we never call one.
Scored Sun Nov 15 · 4:33 PM ET · run 2026-10-reprice-2fb9df4b
Scored Sun Nov 15 · run 2026-10-reprice-2fb9df4b