Every chance we publish is saved before the game starts, then graded from official stats. Nothing is deleted, edited or hand-picked: wins and losses both stay.
Verified, pick by pick
When we favor a side, it hits 57.5% of the time. We said 58.2%.
21,257 favored picks across NBA, NCAAB, NCAAF, NFL, NHL, graded since September 29, 2026.
A chance is only worth something if it's honest. Each row groups picks by the chance we gave them: the bar is how often they actually hit, the white tick is what we said. The closer they line up, the more you can trust the number on a pick.
Each pick shows when it was saved, before its game started, and how it finished. Losses included.
A pick's chance, line and price are stored the moment they're published, with a timestamp. Anything saved after a game starts is never counted.
Results come from the leagues' official box scores, automatically, usually the next morning. A player who doesn't play is voided, not counted as a win.
No pick is deleted, edited or left out after the fact, including the misses. The numbers above are every graded pick.
Download every graded pick, with its timestamp, chance, line, price and result, and run the numbers your own way.
Three bots play the daily card every day with fake money, one per card mode. Each has $10000.00 to start (raised from $1,000 on November 18 for more volume; earlier results still count), uses the Moderate profile, spreads each day's budget across about 30 bets (a card at 11am Eastern, topped up every hour until 10pm as the evening's lines come out) and plays +EV entries of up to 6 picks, flex included, that win at least 10% of the time, on Underdog and PrizePicks. Until November 21 they played 2-pick entries only, which left most days with 5 or 6 bets. The record restarted on November 16, 2026 with model 2.2, which prices every pick differently; the bots' earlier entries were picked by the old model and are no longer counted. Every entry is saved before the games at the listed payout and settled from box scores with the apps' own rules.
Against the sportsbooks' closing price (20 picks): the card gave them 57.1%, the books closed at 52.0%, and 60.0% hit. The books had 55.5% when picked; the close moved toward 10% of 20.
Forecast: entries pay more than their stake 32.1% of the time; actual 31.3%. Expected +$42.29 on settled entries; made +$28.50.
Against the sportsbooks' closing price (22 picks): the card gave them 57.2%, the books closed at 52.4%, and 50.0% hit. The books had 55.2% when picked; the close moved toward 14% of 22.
Forecast: entries pay more than their stake 33.3% of the time; actual 22.2%. Expected +$46.76 on settled entries; made −$11.50.
Against the sportsbooks' closing price (29 picks): the card gave them 57.1%, the books closed at 52.4%, and 55.2% hit. The books had 55.1% when picked; the close moved toward 21% of 29.
Forecast: entries pay more than their stake 33.1% of the time; actual 23.8%. Expected +$48.73 on settled entries; made −$94.00.
Paper trading with fake money, for measuring the model and the card modes on games that hadn't happened yet. Results can swing a lot over weeks; a few months of daily cards say far more than a few days. Not betting advice. 21+.