SHA-256 OF THE SHEET BELOW
299b3e82efe3cd9bfb0d94d2dc8404780974cd266e44bc349598961c86f3b8f1
Posted publicly at the same moment this sent, and read aloud on camera later today. It cannot be changed afterwards, which is the entire point.
BEST BETS
Denzel Boston · Rec yds OVER 33.5 · -115 Model 47.0 · win 71.9% vs market 50.2%
WORTH A LOOK
ATL @ NO · Total UNDER 48.5 · -115 Model 45.8 · win 58.2% vs market 50.8%
THE HONEST RANGE
10,000 runs of this exact sheet. A typical week lands at +23.53u. The best 5% reach +49.10u. The worst 5% sit at −2.81u. The week loses money 7% of the time. A losing week is not evidence the model is broken, and a winning one is not evidence it works. Every pick above gets graded on Tuesday — wins, losses, and the ones the board said to skip — and the grading goes on video at the same size as the wins.
THE WHOLE SHEET
All 234 picks and all 784 refusals, every game, every number — the file the hash above is taken from:
Check it yourself: run that file through any SHA-256 tool and the result has to match the hash at the top of this email, character for character. If it doesn’t, we changed the sheet after the fact and you should stop reading us.
WEEK 1 WAS GRADED WRONG
Two errors in the Week 1 grading. We found them, named every pick they touched, and published the corrected best-bets record — 16-19-1, -7.82u, which is 2.47u worse than what we first put out. The page says exactly what we got wrong:
WHERE THIS GOES NEXT
The board replays every game 10,000 times, reads the data back to 1999, and backtests against seven seasons. It looks for as many different ways to reach the Vegas number as it can — top down, bottom up. NBA and college basketball come next, and we sharpen it every season from the one before.
Four videos a week now — Thursday’s commit, Saturday’s explainer, Monday’s look ahead, Tuesday’s grading. Nobody else shows you what they skipped. Follow along on YouTube and X, and forward this to one person who’d want to see a record kept this way.