NTDevLab / ODP

About ODP

ODP (OpenDota Predictions) is a machine-learning project for predicting the outcomes of professional Dota 2 matches.

How does it work?

The system analyzes large volumes of data from past matches, including:

  • Head-to-head history of the teams
  • Individual player statistics
  • Popularity and effectiveness of heroes in the current meta
  • Synergy between heroes within teams
  • Counter-picks and draft advantages
  • Data about the current game patches

Based on this data, our machine-learning algorithm produces predictions for the likely match winner and other statistical indicators.

What makes it different

Open validation

Every prediction is published before the game and checked against the final result — wins and misses alike.

Machine learning

Models trained on tens of thousands of professional matches, retrained as new data arrives.

Fresh data

Predictions are recalculated when new match data comes in, and the version history stays public.

Independence

Objective predictions based only on statistical data — no narratives, no fan bias.

Our prediction tiers

We use a tiered approach to estimating match outcomes:

  • Base predictions

    Available to everyone

  • High-confidence predictions

    A special group of matches with the strongest signals

  • Premium predictions

    Extended analytics on top of the base model

A prediction is a probability estimate, not a guarantee. Check the open accuracy statistics.