We have some pro dota2 teams matches data, we want to predict some match result using these data, do you have any idears? Give it a try!
Under /ref folder there are 3 files:
- teams_id.csv : the id of pro teams
- heroes_id.csv : the id of heroes (characters in dota2)
- Top_1000_teams_mathces.csv : matches data of top 1000 pro teams, containing both features and labels.
The features are:
Radiant_team_id
: team id of radiant sideDire_team_id
: team id of dire sideRadiant_team_rating
: team's rating of radiant side (offical rating)Dire_team_rating
: team's rating of dire side (offical rating)Radiant_team_hero1-5
: heroes that radiant pickDire_team_hero1-5
: heroes that dire pickstart_time
: time that the match start
The labels (targets) are:
Radiant_win
: if radiant side win the matchRadiant_fb
: if radiant side get the first bloodRadiant_10kill
: if radiant side first get 10 kill
Our model only reach 55% acc on the dev set, a little better than random guess... Hope you can do it better!