Originally published 29 May 2002

Risk researchers at the University of Warwick have been commissioned by Radio Five to produce a statistical model designed to predict the odds of every individual match in the World Cup. The model has been dubbed the Glover Automated Results Indicator (GARI), after Fi Glover the Radio Five Live presenter who introduced the model on her show (22:00-1:00 on Tues 28th May 2002). Updates of the model and its performance will be available on the Five Live web-site www.bbc.co.uk/fivelive.

The model takes in all the qualifying, play-offs and friendly games (around 900) for all the participating teams (around 200) to generate ratings for each team. Each team is rated on two dimensions – ‘Strength’, which quantifies how the teams measure up against each another, and ‘Patchiness’, which charts the unpredictability of a team. These estimates are then further adjusted using data from FIFA and bookmakers. Finally the model is used to run a “Monte Carlo” simulation of the entire tournament. The analysis addresses questions such as:

Forecasting: Who are the best teams and the odds of their winning their groups, getting to the quarter finals, and so on?

• England has a 62% chance of qualifying from Group F, but still faces a 20% chance of losing to Sweden.
• Argentina is GARI’s favourite to win the Cup, with a 13.2% chance, whilst France, the bookie’s favourite, lies in third place at 8.7%. England’s odds are estimated to be 6.9%

Red Flagging: Which teams have the highest ‘upset’ potential, i.e. the ones with the patchy record of ignominious failure peppered with occasional giant slaying?

• GARI predicts several possible upsets in the first round including Senegal beating France, South Africa beating Spain and Mexico beating Italy
• Home team advantages are likely to place Japan at the head of Group H and Korea second in Group D

The model was created privately by Henry Stott, a director of the Financial Decisions Research Group (FDRG) at the University of Warwick www.warwick.ac.uk/psych/.

GARI was originally built to help win an office sweepstake. However it is also an interesting case study on the strengths and weaknesses of artificial intelligence compared to human reasoning. On the one hand, GARI cannot replicate the wide ranging but uncomplicated calculations achieved by human counterparts. On the other hand, the analysis integrates a narrow range of data with a depth and complexity that could never be achieved by a human mind. GARI’s ability to then impartially simulate the thousands of permutations of how the tournament might develop is another task no human could replicate.

For further details please contact:
Henry Stott, University of Warwick
Tel: 024 7657 3127
Email: h.p.stott@warwick.ac.uk