NTDevLab / ODP
ODP Project — ML predictions for Dota 2 matches
An independent ML model for professional Dota 2: match winner, total kills, hero synergy and counter-picks. Predictions are published before the match starts and stay in the open history — mistakes included.
- Python
- Flask
- ML
- PostgreSQL
- Redis
- Active project
- Interface languages: RU
The problem it solves
Dota 2 markets are hard to evaluate manually: drafts change daily, hero pools shift patch to patch. ODP turns match data into explicit probabilities — before the game starts.
Data and ML
The model works on professional match statistics: picks and bans, hero win rates, synergy and counter-pick matrices, kill totals. The ML pipeline is maintained by NTDevLab.
The NTDevLab approach
The prediction is published before the event and is never rewritten afterwards: the open history keeps both hits and misses, so model accuracy can be checked independently.