The Leicester Fluke – What went wrong with the Predictive Model

The 2015/16 English Premier League finished off with an unlikely champion. Leicester City football club emerged victorious this season despite their not-so impressive record in the past. In a podcast episode of More or Less Tim Harford explores why Leicester City’s win came as a surprise.

Courtesy Peter Woodentop and Wikimedia Commons
Leicester City Football Club after their unpredicted victory in the Premier League

No one had quite predicted it right, not even Joe Prince Wright, a lead writer in NBC Sport Soccer Website. Tim Harford tells in his podcast that Wright has one of the most accurate football prediction records, but failed to predict Leicester City’s achievement and neither did his colleagues at NBC.

Is the number of goals made by the each team enough to tell who’s going to win? If this were the case, then they should have gotten it right. Sports analysts were caught off guard with Leicester City’s win. What went wrong with the predictions? There are actually 2 ways the most reliable football predictors failed.

Impressions were based solely on the number of goals and shots on target

In Lord Danny Finkelstein’s Fink Tank predictor, result simulations were according to the teams’ goals and shots on target that were included in their model. Using the simulation, analysts can predict the outcome of the whole football season through the ranking of different teams. The simulation is not a one-time process. It has to be repeated a number of times to identify the differences of results.

Leicester City had shown minimal performance in past seasons. Based on the factors that affect the Fink Tank simulations, it is very unlikely for the team to be included in the top 5. Teams outside the top 5 had a 1.1% chance of winning the season. The predictor did not consider the team’s improvements in their games this season, and failed to give the analysts insightful changes of rankings.

The group’s performance was underestimated

There was not much to expect from Leicester City, or so they thought. There was quite an improvement in their performance this season that eventually brought them to the championship. As Lord Finkelstein put it, the best team does not always win, because in this team’s case, it is a combination of luck and improvement.

There were years that the team played so badly that most analysts thought that they would come out somewhere in the middle on the ranking on the Premiere League at best. However, their game changed dramatically and showed a really good defense, outperforming the other teams. The number of injured members from the opponents were also an advantage of Leicester City’s victory.

Football forecasting is something that fans look forward to every football season. From the most unusual predictor like Paul the Octopus to a more reliable one like Fink Tank, football fans crave data projections that would help them anticipate the results in the championship.

Before, football forecasting was about counting the number of goals per game and that’s how the insights were gathered. However, the unpredictability of a team’s overall performance was not taken into account. A good forecasting shouldn’t be linear and dependent solely on past actuals. Modern analytics tells us that the quality of the performance and the number of shots in and out of target should also be considered.

One thought on “The Leicester Fluke – What went wrong with the Predictive Model”

Comments are closed.