Azubu Prediction Engine
Leading up to the 2015 League of Legends World Finals my team built a program that would predict the winners of each match.
Graphic Overlays would be displayed within the Azubu player prior to the start of each match. They would display the chances each team had of winning the game. Players were given Individual Scores that showed how much their performance could influence the outcome of each match.
Our prediction algorithm utilized a database of every competitive game from 2015 LCS Season. It pulled JSON data via the Riot API and utilized billions Random Forest Decision Trees to predict a winner. Our picks outperformed 84% of the participants Riot's "World's Pick Em" promotion with an overall accuracy 64.58%. Only 1 Riot caster beat APE's predictions in 2015.
