A virtual gambler developed by Australian computer researchers to predict the outcome of team sports has beaten human expert tipsters and could see bookies getting fleeced. The computer eschews emotion to rely only on statistical data to pick winners, analysing factors such as a team's win-loss record, ranking and home ground record. Alan McCabe of James Cook University, in Queensland, claims his virtual tipster 'MAIT' - McCabe's Artificially Intelligent Tipper - picked the winner in more than 80 per cent of matches in the 2000 Australian National Rugby League (NRL) competition. MAIT - which Mr McCabe said could potentially be converted to cover other team sports such as soccer - was unveiled at a conference in Canberra last week. The computer put four human tipsters to shame when it correctly predicted the winner in 151 of 182 NRL matches in the 2000 season. By comparison, the most successful of the humans picked the correct result in only 120 matches. Using odds from an Australian betting agency, the researchers calculated that betting on the computer tipster's recommendations would have returned a profit of between 39 and 46 per cent, depending on strategy used. News of MAIT's success comes after the SAR government announced last month that soccer gambling would be legalised next summer, issuing a five-year licence to the Hong Kong Jockey Club. Mr McCabe said on his Web site that MAIT did not know how to take into account subjective factors such as the impact of an injury to a star player. 'From a financial point of view it would be advantageous to examine more advanced betting strategies to maximise return and minimise the number of bets lost,' Mr McCabe said. The research into soccer gambling is a sideline from other work using a 'neural network' computer system - capable of a limited form of 'learning' - to analyse handwriting.
