140 likes | 616 Views
League of Legends:. Predictive Analysis on the Professional Scene By Justin White. Background. What is LoL? 27 million daily players worldwide 67 million monthly players 5 vs. 5 118 champions currently. Picks and Bans. Prior to game Ban three champions Draft champions.
E N D
League of Legends: Predictive Analysis on the Professional Scene By Justin White
Background • What is LoL? • 27 million daily players worldwide • 67 million monthly players • 5 vs. 5 • 118 champions currently
Picks and Bans • Prior to game • Ban three champions • Draft champions
Professional Scene • Primary Tournaments for each Region • North America • NA League Championship Series • Europe • EU League Championship Series • Korea • OGN
International Tournaments • Season 3 World Finals • 2 million dollar prize pool • 14 teams
Problem Definition • Predict what champions one team will pick • Given variable number of bans • Given 5 champion picks • Given time • Output 5 picks • Dynamic Bayes Networks • Relates variables over timesteps
Implementation • Generate MySQL Database • Fill with picks and bans • Each champion is given a number • Team 1 picks… • Team 2 picks... • Time • Win/Loss
Implementation • Projeny • The Probabilistic Networks Generator in Java • Requires • Matlab • Bayes Net Toolbox • JMatLink http://projeny.sourceforge.net https://code.google.com/p/bnt/ http://jmatlink.sourceforge.net/
Status • Database • Environment • Test learning algorithms • Bayesian Parameter Learning • Batch MLE/MAP parameter learning using EM