Model breakdowns, edges, and the thinking behind the numbers.
Looking for the daily email editions? View Newsletter Archive →
Plus The Morning Board daily: every model graded in public, free. No spam, unsubscribe anytime.
The Brier score is how we rank every AI model on our leaderboard, and it's friendlier than it sounds. Here's what it measures, why it beats win-loss records, and what a good one looks like.
Every AI model and tracked expert on our board now has a head-to-head page against every other: same games, graded finals, and who was right when they disagreed.
We graded 180 games where every AI model picked the same team. Unanimity underperformed a two-model agreement, and last Monday it went 2-5. Here's why consensus fails.
Sol Pro and Luna Pro posted near identical records on our board for two months. One costs fifty times the other. Here is the math behind the first retirement of a healthy model.
Grok 4.5's over picks hit 37% for a full month. We stopped asking it to be right and started betting the other side. Here's the full graded record of the strategy.
The UFC's first trip to Serbia is an afternoon card in the States. We break down every main card fight, what twelve AI models predict for each, and where the field disagrees with the betting market.
Hey everyone. Fight week is here, and this one lands Saturday from the Etihad Arena in Abu Dhabi. Magomed Ankalaev headlines his first walk to the Octagon since losing the light heavyweight belt to...
Two former champs trying to get back in the saddle, an undefeated prospect against a 29-15 veteran, and a 42-year-old walking into a buzzsaw. We ran the whole main card through the model and asked nine AIs what they think.
The first two rounds of a 12 team half PPR draft, built from five projection systems, six draft markets and de-vigged Vegas props, with the receipts on every call (including the round three tight end everyone keeps letting slide).