Predicting the Premier League finish with extensive data modelling

0
49

Football has returned! This last weekend saw the Bundesliga resume play and strong television ratings – with double the average audience in Germany and FS1’s largest ever audience in the United States! Now eyes will turn to the remaining big European Football leagues (Premier League, Serie A and La Liga) to see when (or if) they will return.

Consider the case of the Premier League. They have been negotiating their return for several weeks with an eye on a June restart. This has received pushback from both players and several clubs. Even more interesting, the Scottish Premiership this week unanimously decided the season could not be concluded and simply awarded Celtic the title.

Part of the challenge to restarting the season is the resistance of six clubs to playing the remainder of the season based on the fairness of completing the season under different circumstances and the economic result of relegation on these teams. After the work DataRobot has done predicting the Champions League knock-out stage, the French Ligue 1 and Bundesliga conclusion based on how the teams were playing prior to the postponement of the season, I was curious if we could determine how the Premier League season was likely to conclude under normal circumstances.

We used the same models and approach we developed to predict the knockout stage of the Champions League, using the game data provided by Data Sports Group. We used DataRobot’s models to predict which team would win each match to simulate the final nine matchdays 10,000 times. For each team, we calculated the average number of wins, draws and losses over those 10,000 seasons to build an average final table.

In the remaining nine games, we see few changes of team standings. Liverpool continues its historic run to the title and adds to the historic nature of the season by tying the 2017-18 Manchester City club for most points in a Premier League season with 100. Liverpool will beat that club with a winning point margin of 23. Manchester City will hold onto second place, with the three additional Champions League Group Stage positions going to Leicester City, Chelsea and Manchester United.

Wolverhampton and Sheffield will hold onto the Europa League positions, while Tottenham and Arsenal will tie just three points from the Europa League slots. Brighton & Hove Albion, West Ham and Watford will cling to their Premier League spots for next year, with West Ham and Watford tying just one point above Aston Villa. While Aston Villa falls just one point from remaining in the Premier League it does pull ahead of Bournemouth as Bournemouth, Norwich City and Aston Villa all face relegation to the Championship.

It will be interesting to see, when/if the season resumes how things play out. Will we see a record tying or breaking year from Liverpool? How will teams perform in empty or neutral sites? What will be the impact of several months off on the injury situations for these teams and how will that impact the final standings. We are looking forward to finding all of this out.

Contact our Sales Team to learn how you can apply predictive analytics, machine learning and AI for your sport business.