Every call we publish is scored against what happened, including the misses. The call scored for each player is the final call we made on the player before his kickoff. The rule changed on 2026-11-01. Under the first call we made, the season's record read 66.9% over 2791 graded player-weeks. Only weeks that have been scored are shown here.
We were not live in week 1
We made no call that week, so there is nothing of ours to score. A week that did not happen is not a week we got wrong.
67.0%Week 3 · settled258 calls · 70.4% when confident
83.8%Week 4 · settled314 calls · 100.0% when confident
70.2%Week 5 · settled359 calls · 92.9% when confident
71.6%Week 6 · settled377 calls · 92.6% when confident
73.3%Week 7 · settled386 calls · 96.0% when confident
85.0%Week 8 · settled326 calls · 96.8% when confident
76.6%Week 9 · settled401 calls · 95.1% when confident
79.5%Week 10 · settled370 calls · 96.0% when confident
75.9%season accuracy2791 calls of ours, settled weeks
43 h aheadmedian head start · seasonmiddle half 35 h ahead to 48 h ahead · 816 players
Head start is the hours between our first call on a player and his club's own word on him. It is counted only for players the club spoke on.
Each of our five bands, tested against what happened, over 2791 calls of ours in settled weeks. A band is a range of the chance we gave that he plays. This checks the chances themselves, not whether a play or sit call was right.
Each band: its calls, how many played, the rate we published it at, and the rate it showed
Band
Calls
Played
Published at
Showed
NOT EXPECTEDwe do not expect him to play
74
1
2%range 1–2%
1.4%95% range 0.2%–7.3% · not enough calls to judge
UNLIKELYunlikely to play
129wider book 3 · designated-player book 126
10wider book 0 · designated-player book 10
not measured in this bookwider book · designated-player book 22%, range 19–25%
0.0%wider book, 95% range 0.0%–56.1% · not enough calls to judge · designated-player book 7.9%, 95% range 4.4%–14.0% · all books together 7.8%
COIN FLIPa genuine coin flip
554wider book 138 · designated-player book 416
179wider book 18 · designated-player book 161
34%wider book, range 31–38% · designated-player book 53%, range 52–55%
13.0%wider book, 95% range 8.4%–19.7% · designated-player book 38.7%, 95% range 34.2%–43.5% · all books together 32.3%
LIKELYlikely to play
694wider book 283 · designated-player book 411
408wider book 95 · designated-player book 313
76%wider book, range 75–78% · designated-player book 75%, range 74–76%
33.6%wider book, 95% range 28.3%–39.3% · designated-player book 76.2%, 95% range 71.8%–80.0% · all books together 58.8%
EXPECTEDwe expect him to play
1340wider book 1052 · designated-player book 288
1226wider book 1002 · designated-player book 224
96%wider book, range 96–97% · designated-player book 87%, range 85–88%
95.3%wider book, 95% range 93.8%–96.4% · designated-player book 77.8%, 95% range 72.6%–82.2% · all books together 91.5%
Share who played (dot) and its 95% range (bar), 0–100% · the wider book, where a band holds its calls
NOT EXPECTED1.4% range 0.2%–7.3%
1 played of 74 · published at 2% · not enough calls to judge
UNLIKELY · wider book0.0% range 0.0%–56.1%
0 played of 3 · not measured in this book · not enough calls to judge
COIN FLIP · wider book13.0% range 8.4%–19.7%
18 played of 138 · published at 34%
LIKELY · wider book33.6% range 28.3%–39.3%
95 played of 283 · published at 76%
EXPECTED · wider book95.3% range 93.8%–96.4%
1002 played of 1052 · published at 96%
The same for the designated-player book, against its own published rate
UNLIKELY · designated-player book7.9% range 4.4%–14.0%
10 played of 126 · published at 22%
COIN FLIP · designated-player book38.7% range 34.2%–43.5%
161 played of 416 · published at 53%
LIKELY · designated-player book76.2% range 71.8%–80.0%
313 played of 411 · published at 75%
EXPECTED · designated-player book77.8% range 72.6%–82.2%
224 played of 288 · published at 87%
The score, split in three
Brier score 0.1587 = miscalibration 0.0204 − discrimination 0.0881 + uncertainty 0.2264, over 2791 graded calls. Lower is better for the Brier score and for miscalibration; higher is better for discrimination.
The three parts of the Brier score
Part
Value
Range
Uncertaintyhow hard the question is, whatever the forecaster
0.2264
Miscalibrationhow far our numbers are from what happened (0 is a number that means what it says)
0.0204
Discriminationhow well our numbers separate who plays from who does not
0.0881
90% range 0.0806 to 0.0988
Skillour Brier score against always saying the share who played (below 0 is worse than that)
0.2992
90% range 0.2623 to 0.3341
Settled weeks only. The full analytics, week by week and call by call, are for signed-in accounts.