Nothing has resolved yet, so there is no score. When contracts settle, this shows our Brier against the market’s on the same contracts — which is the result that matters, not the profit column.

Every bet is rated out of 100 from three measured things: how far our number sits from the price once our own bias on that market is removed, our demonstrated skill over a naive baseline on that kind of contract, and how tight the spread is. Skill multiplies rather than adds, so a market we cannot actually price — walks allowed, where our measured skill is −0.0006 — cannot earn a good rating by showing a big edge. The big edge is the symptom. Every market and the full record.

Win probability is anchored to what really happened in the same situation across 9.3 million plate appearances, then adjusted for the two teams playing. Scores come from simulating the rest of the game thousands of times. Both are estimates and both are wrong sometimes; the page tries to be clear about how often.