How it works

The project is in beta test. Further updates are coming.

The predictor is neural network trained on competitive history of Dota 2 matches. Updates after every match. Performance depends on available statistics. China tournaments show best results so far.

What is considered

  1. Overall players’ strength
  2. Pick vs pick advantage/disadvantage
  3. Team performance trends
  4. Teams heads-up stats

What IS NOT considered

  1. Current motivation of a player (play-offs/group stage/tournament importance)
  2. Temporary replacements are not treated well yet
  3. Leagues without Dota 2 TV tickets are not used for training