Dota 2 match
predictions you
can trust
19 machine-learning models break down team form, drafts, hero meta and Elo ratings — and give a win probability for every match. Pro and amateur tournaments, updated in real time.
From raw data to a win probability
No magic, no gut feeling — just match data, rosters and models that learn from history.
Data
Live matches and history from OpenDota, current rosters and brackets from Liquipedia. Refreshed automatically every few minutes.
Models
An ensemble of 19 models: Elo and H2H ratings, recent form, a draft XGBoost, hero-meta strength — and their weighted consensus.
Prediction
Each team's win probability with a model-consensus view. You see where models agree and where a match is closer to a coin-flip — no false confidence.
Model families
Every prediction is the agreement of several independent approaches. Here are the main ones.
Elo & H2H rating
A dynamic team-strength rating and head-to-head history.
Recent form
Weighted recent results — who's on a run and who's slumping.
Draft XGBoost
Gradient boosting rates the pick/ban phase and hero synergy.
Hero meta
Current win rates and hero priority in the live patch.
Ensemble blend
The weighted consensus of models — the final match probability.
Real-time updates
Predictions recompute as rosters and results update.
Predictions for today's
matches are ready
Open dota2predict.gg — win probabilities, model consensus and the match schedule. Free.
Go to dota2predict.gg →