Disagreement Lab

Upcoming fights where the model, the betting market and the fans see it differently, biggest gap first. That is where the reasoning is worth reading and where value bets come from. Part of full access.

Overview and filters

Three signals give a win probability for every fight. Model is our machine-learning model. Market is the bookmakers' price with their margin removed. Fans are FIGHT//AI users' votes once a fight has five or more, otherwise social-media sentiment. A signal with no data, or fans sitting on 50/50, is left out rather than counted, so a fight nobody has voted on is not a fake disagreement.

The score is the biggest gap, in percentage points, between two signals for the same fighter. High is 15 points or more, Medium 8 to 15, Low under 8. The tiles count fights at each level and double as the filter; the chips narrow by which two signals clash (Model vs Market, Model vs Fans, Market vs Fans) and by event, and the search box finds a fighter or event.

Disagreement card

The badges give the level with the score, the pair that is furthest apart, and Different winners when those two back different fighters. Both fighters sit at the ends of the signal track: the left end is 100% for the fighter on the left, the right end 100% for the one on the right, and the middle tick is 50/50. Each signal is a coloured dot, cyan for the model, orange for the market and violet for the fans; the band joins the two furthest apart.

The sentence says the gap in words. The tiles give each signal's favourite and its %, with the source of the fan number. When the model is at least five points higher than the market on a fighter, a note says so: that is the kind of gap the AI Picks bot looks for. The whole card opens the fight page, where the Fight Map, the market panel and the news explain the split.