Communications of the ACM
Software to Predict 'march Madness' Basketball Winner
Courtesy of NCAA
The goal of the March Madness Predictive Analysis Challenge, now in its second year, is to build software that can select winning basketball teams with greater accuracy than humans.
Machines compete against machines to determine which algorithm can make the most accurate win predictions. Tournament brackets must be selected wholly by algorithm, and no specific team-based rules are permitted. All contests are limited to using the same data set, comprised of team and player statistics from the 2006 season until February 2011.
Contest organizer Danny Tarlow has devised an algorithm that sifts through volumes of regular season data and applies probabilities toward the finding of equations that match the outcomes of each game. The software then employs those equations to choose which teams will win.
"The algorithm ... just sees the outcome of each game in the season, and it tries to discover latent characteristics that best explain the outcomes," Tarlow says. Other contest entries include using genetic algorithms programmed to evolve equations capable of choosing winners, and software designed to abstract a team's strengths and weaknesses into a single number, then select the team with the higher number in each game.
From New Scientist
View Full Article
Abstracts Copyright © 2011 Information Inc. 
, Bethesda, Maryland, USA
No entries found