Å®ÉúСÊÓƵ

The FIFA Women’s World Cup kicks off today in France, with 24 teams competing for football’s biggest prize. While there have been many attempts to use statistical models to predict results in men’s football, the women’s game has received less attention from statisticians so far. , led by Andreas Groll at TU Dortmund University in Germany, has used machine learning to create a

The model combines three state-of-the-art forecasting methods. First, it estimates each team’s ability based on results from the last eight years, with more recent results weighted more heavily. It generates a second ability estimate for each team based on odds from 18 bookmakers, taking into account their profit margins.

Finally, it uses machine learning to combine the two ability estimates along with a broad range of other data, including FIFA rankings and socioeconomic factors, such as the country’s population and GDP, as well as which team is playing at home. Some of these factors are more relevant than others, but the model learns which are most important by comparing the data with results from the last two World Cups.

The model estimates how many goals each team will score in every possible match, allowing it to compute the probability of a win, draw or loss. Using these probabilities, the team simulated the entire tournament 100,000 times to determine every team’s chances of progressing to each stage.

The results have USA as clear favourites, with a 28.1 per cent chance of lifting the trophy. France, England and Germany are some way behind, with chances of 14.3, 13.3 and 12.9 per cent respectively.

This contrasts with the bookmakers’ assessments, which give USA and France a roughly equal chance of winning the tournament. This may be because France are playing at home, but the statistical model gives home advantage little weight: the hosts of the last two World Cups went out in the quarter finals.

So is it worth putting some money on a USA victory? No, say the researchers. The bookmakers are offering less generous odds, relative to actual probabilities, compared with the men’s World Cup, perhaps reflecting less confidence in their own predictions.