FAQ

Where do match predictions come from?

20 independent AI models, each with a different signal (Elo, recent form, hero meta, draft, etc.). They run automatically; the ensemble forecast on each match is their mean.

Where does the data come from?

Match results, drafts, and team IDs from OpenDota. Upcoming schedules and rosters from Liquipedia. Both are public sources.

Why is model accuracy showing 0% / "Calibrating"?

We only count accuracy on matches the models predicted before the result was known. Already-finished matches that landed in our DB later are excluded — backfilling them would inflate accuracy (the models would have seen those games during training). Numbers populate as upcoming matches complete.

Why is the prediction pool empty / odds showing "—"?

The pool is pari-mutuel — winners split the losers' pool. With $0 in the pool there's nothing to split yet, so we don't show a multiplier. Once someone bets, odds appear.

What's a "TBD" match?

Bracket placeholder — winner of a previous match. We list them so you see the schedule, but no models or predictions until the team is determined.

Why does NaVi sometimes play "HULIGANI" here but "L1GA TEAM" on other sites?

OpenDota stores teams by canonical name; some Russian-language sites use a sponsor or league nickname. Same team, same match.

Is this real-money betting?

No. The wager system uses internal site balance you load via crypto. Pari-mutuel, no bookmaker margin. 5% commission on the winning pool. Predictions are Pro-only.

Do I need a separate account from cs2predict?

No — same account, same balance. Sign in once.

I found a bug.

Open a ticket from Support if signed in, or email hi@dota2predict.gg.