Every prediction is graded against the real final score once the game ends. Nothing here is cherry-picked — these are the same numbers we look at internally, updated automatically as games settle. Results are broken out by the confidence tier each call was made at, because a model that can't tell you when it's guessing isn't worth much.
"All Sports" excludes CS2 from these figures — CS2 roster composition varies too much match-to-match (frequent stand-ins) to track consistently against a fixed prediction. Select CS2 above to see its numbers directly. ATP is excluded for a different reason — it's a brand-new engine with no settled track record yet; it'll fold into the blend once it has one. Select ATP above to see its numbers directly as they build up.