Scorecasting prediction model back for another season
09 August 2019
The computer programme for predicting football scores developed by University of Reading economists is back for a second season.
RED, the computer modelling programme that took on Mark Lawrenson for BBC’s Football Focus last year, has begun making its predictions for the football season including final places in the Premier League.
The model has already produced predictions for the first round of the English Football League, successfully forecasting that Championship newcomers Barnsley would defeat Fulham, who many reckon will fill one of the automatic promotion spots next summer.
Dr Carl Singleton, Lecturer in Economics at the University of Reading, said:
“Last season, RED took on the BBC’s Mark Lawrenson on Football Focus and won, casting fresh doubt on man’s supremacy over the machines. This season RED has been re-booted and upgraded, providing many more and more detailed predictions on football match outcomes, hoping to prove that a bit of forecasting science and computer power can do better than the judgements of professional football pundits and fans alike.”
The most likely score in each game is calculated using a complex metric that analyses historical results for each team, as well as things like form, league position and even the day of the week the match is played. This gives an estimated number of goals for both sides in any given fixture, allowing the most likely result and score to be given.
Tests comparing actual results from previous years with what would have been predicted by the model, as well as the model’s experience last season, shows it is competitive with, and often beats, the predictions made by experts (i.e. ex-professional footballers and managers) over the course of a season.
Dr James Reade, Associate Professor in the Department of Economics at the University of Reading, said:
“With just days to go before the Premier League season begins, fans up and down the country are excitedly wondering how their team will fare, and if this could be their year.
“We have updated RED to try to provide even more accurate predictions and he/she (we aren’t sure) certainly gives some fans early reason to cheer or need cheering up.”
Find out more about the model and see more predictions at http://blogs.reading.ac.uk/econscorecast/
The predictions produced are what the model suggests are the most likely outcomes and show the chance of these occurring as a percentage (rounded to the nearest one decimal place). This season the model is also outputting forecasts of the expected goals in matches by each team.
It should go without saying, these predictions can never be 100% accurate, far from it, and just provide a rough guide to the most likely results, which can be influenced by several factors, such as team selection, and unpredictable events during a match. The University of Reading does not promote gambling and does not intend for the predictions to be used as a guide for this purpose.