from Hacker News

Ask HN: Forecasting FIFA World Cup outcomes, statistically, for fun

by otto_ortega on 6/7/18, 9:28 PM with 1 comments

We are 1 week away from FIFA World Cup 2018, given than it is one of the most important sport events worldwide I though it would be an interesting and entertaining exercise to try and predict the outcomes of the matches, Im wondering what are the most effective approaches to do so and if there are some free data sources out there that one can use to build a model.

For predicting the odds of a team winning an specific match it seems that there are 2 approaches that have been proved to be effective:

- Using Elo Ratings [1] - Using the bookmaker consensus model [2]

And we can easily obtain data to setup a model based on those approaches from the Elo Rating website and betting odds comparison sites respectively.

Now, for predicting the score of a given match it seems most approaches seems to use statistics about shots per game, shots on target, and goals per game. But I couldn't find any website to use as data source for that information.

So Im wondering what models or techniques could the HN community suggest, and if someone is aware of some website to use as data source to create a model to predict a match score.

[1] http://www.eloratings.net/about [2] https://arxiv.org/pdf/1710.02824.pdf

  • by marketgod on 6/7/18, 9:40 PM

    You can track the movements of the odds for already scheduled games and look for movements in the odds/spread that are unusual.

    Off topic, but to make money:

    You can open an account with a book in both home countries, and try to bet the home team on the away book and visa versa. Most locals bet on their home team so if you bet in the opposite country, your odds get better.