Who says cricket matches are unpredictable? A business solutions company in Mumbai has devised a computer model that forecasts cricket scores with apparently astounding success.

The company, Fractal Analytics, scored a major hit by predicting West Indies' underdog victory over Pakistan in the opening game of the World Cup on Tuesday.

Amazingly, it even forecast the exact scores of rival captains Brian Lara of the West Indies and Inzamam-ul-Haq of Pakistan.

Tuesday morning's edition of the Mumbai-based Daily News and Analysis newspaper carried the predictions, based on what what the company said was scientific analysis of players' past performances.

The forecast said: "West Indies will win in a tight finish and will put up a low score if they bat first."

The West Indies, batting first, managed a moderate 241-9 after being 183-6 at one stage.

Pakistan lost the game by 54 runs but were cruising at 99-3 and fought until the end through Shoaib Malik's aggressive 62.

The prediction correctly said Lara would make 37 and Inzamam 36.

Fractal Analytics, which specialises in predictive analysis of customer behaviour for clients such as Visa and Citibank, devised the model to forecast cricket results just for fun.

"We started this a few months back and found that the predictions actually came out true," Mukesh Budania, a consultant with the firm, told AFP by telephone from Mumbai.

"So we decided to be bold and give our predictions to a newspaper a day before the match.

"I can tell you the whole team in the office was jumping with joy last night when Lara got out for 37 and Inzamam for 36."

The model is based on the theory that the outcome of every game and the performance of every player depends, to a degree, on what the player or the team has done in the past.

To predict which team will win a particular match, the statistics of all one-day internationals played since 1996 are looked at to identify factors that have the maximum impact on the outcome of the match.

Factors found to be relevant include the relative forms of the teams, their recent past record and their performances against each other.

Based on an in-depth analysis of these factors, the model is apparently able to predict not just the result, but individual scores and the winning margin as well.

Budania claims to have correctly predicted results of 70 percent of the matches analysed so far, including two of the three upset win by New Zealand over Australia last month.

Testing the model on results of the last World Cup in South Africa in 2003, the model had a success rate of 75 per cent, he said.

To gauge how a batsman will perform, the model takes into account his performance in the last few matches and looks at which other stages in his playing career the player exhibited similar form.

This data is used to predict the score of the batsman. Other factors like the opposition and the importance of the match are used to refine the predictions.

The big question is who will win the World Cup -- but sadly Budania says it is too early for the computer to tell.

Its next prediction will be for the England-New Zealand match on March 16, which will be published before the game in the DNA newspaper.

"So far we are doing it just for fun, but if the model works we may even think of marketing it in future," he said.

Though it was unclear who would benefit from the model -- fans, or bookmakers.

(Article: Copyright © 2007 AFP)