Statistical model predicts India loss3 min read . Updated: 08 Sep 2007, 12:14 AM IST
Statistical model predicts India loss
Statistical model predicts India loss
The oracle has spoken: on Saturday, India will lose to England in the seventh and last match of the Natwest one-day series, a virtual final, with the series tied at three games each. The margin of defeat in the game that is scheduled to be played at Lord’s: three runs.
The oracle in this case is Score Wizard, a statistical model developed by Mumbai-based predictive analytics company Fractal Analytics Ltd. Data provided by the company on Friday showed that Score Wizard had correctly predicted the outcomes of five of the six matches that have been played in the series thus far. And Srikanth Velamakanni, an alumnus of the Indian Institute of Management Ahmedabad and the Indian Institute of Technology Delhi, claims that during the last Cricket World Cup in the West Indies, Score Wizard’s success rate was 78%.
Fractal, which offers analytics solutions for retail, financial, banking and telecommunications firms to help them cobble together their marketing and risk management programmes, says it’s fairly certain of its findings, as all calculations are based on hard data. “There’s no human judgement involved," says Velamakanni.
In the firm’s reckoning, the critical factor that goes against India (at least in the statistical model) is that the team is playing away from home.
In London, leading bookmaker William Hill, too, believes the hosts will walk away with the NatWest trophy; it is offering odds of 11 on 8 on England, compared to 1 on 1 for India. England are the favourites, says company spokesperson Rupert Adams, explaining the odds: a wager of £1 (Rs82.30) on India would fetch the bettor £2 if the team won; a bet of the same amount on England would yield only £1.70.
The result is different from one predicted by Score Wizard on Friday morning. That said India would win by 30 runs.
It’ll not be a very narrow victory, but it won’t be a comfortable one either, Velamakanni had said in a telephonic interview. By afternoon, his team had come out with a different prediction—that of India’s loss by a heart-wrenching margin.
A simple error had been made when entering the data, explains Rahul Paharia, a me-mber of the cricket predictive model team. Instead of marking England the hosts, the team had entered India.
In that calculation, India had totted up 30 points which, in Fractal’s statistical model, implies a victory by 30 runs. When the correction was made, India’s score dropped to minus three, implying a defeat by three runs.
Paharia says Score Wizard has at least 22 variables. He declines to identify the entire metrics, but says there are six significant variables: performance of a team over the past few matches, performance against the opposition, type of pitch, toss outcome, whether a match is at home or away and, finally, the international ranking of a team. Each category is given a weightage arrived from calculations, and a result on the outcome derived.
Fractal also has a list of variables that made 100% prediction impossible. It calls these “unpredictable noise." Such variables include dropped catches, rain and a batsman getting out on a no-ball.
The firm has another model that can be used to predict the performance of players, although it hasn’t crunched the numbers on these for Saturday’s match. The success rate here is lower, 33%. Velamakanni, who co-founded the company with two other alumni of IIM-A, says the firm does not use this model to predict scores of “unpredictable" players such as middle-order batsman Yuvraj Singh and wicketkeeper-batsman M.S. Dhoni. “We have a problem predicting their performance, but we do very well with Sachin (Tendulkar), Saurav (Ganguly) and (Rahul) Dravid," he says.
Velamakanni says the entire Score Wizard programme began as an experiment and remains that. Fractal does have business plans for the model, though these haven’t been firmed up. The statistical model can be customized for baseball, soccer and American football, he says, and firms that could be approached include betting firms Ladbrokes and William Hill.
Will overseas buyers be interested? William Hill, for instance, isn’t. “We are very happy with our model," says spokesperson Adams.