What the Moneyline picks list actually did. Every pick is recorded before the fight happens — the selection, the strategy behind it, and the closing price — then graded against the result. This is the picks record, not the model's raw hit rate over every fight on the card.
Published picks: 13/18 correct (72%) · +6.8% return (betting the same amount on every pick)
Round total picks: 5/16 correct (31%) · -17.4% return (betting the same amount on every pick)
Reconstructed (not published in advance): 28/32 correct (88%) · +20.8% return (betting the same amount on every pick)
Last 8 events
Moneyline picks use two scheduled checkpoints anchored to UFC's official event StartTime. Before T-24, live picks are published as unconfirmed and may enter, leave, or flip as odds move. At T-24 the current rule/fighter combinations become the maximum eligible set. From T-24 to T-2 that set can only shrink: a removed fighter/rule cannot re-enter. At T-2 the survivors and their reference prices lock permanently and are labeled confirmed. The ordinary market snapshot continues separately until the book pulls the line, preserving the actual close without allowing it to change the pick. Once official results arrive, only the PICKS are graded. Fights the strategies passed on are not counted for or against -- a card where the model was right about eleven fights we never selected is not a good week. Picks with no decisive result (draw or no-contest) are excluded. Return figures assume the same stake on every pick, settled at that pick's own T-2 reference price (closing price on legacy cards) -- so '+8% return' means that for every $100 risked across the picks, $8 came back on top. They are a record of published selections, not advice on stake size. Some older cards are marked RECONSTRUCTED: those picks were not published in advance. They were worked out afterwards by replaying the same five rules over the stored odds, using a model bundle trained only on fights before July 2025 -- so it had never seen the fights it is picking. That is an honest out-of-sample exercise, but it is a weaker claim than a live pick, and the two are never added together on this page. Round total picks (over/under how long the fight lasts) are recorded and graded the same way, in their own table, and reported in their own column. They are never pooled with the moneyline figures: it is a different market, priced by different books at a wider margin, and one combined hit rate would describe neither. A rounds pick is a side AND a line, so a fight quoted at both 1.5 and 2.5 can be picked at one and passed at the other, and each settles on its own.
Historical performance is not a guarantee of future results — see the Disclaimer for the full picture.