This is a guest post by David Sumpter who is currently writing Soccermatics, which will come out on May 5th 2016 with Bloomsbury Books. This blog post is part of one of the book chapters on “Betting against the masses”. You can find out more by following Soccermatics on Twitter.
Before the 2014-15 season, Joe Prince-Wright, wrote down a list of the final positions between one and twenty for each of the teams in the Premier League and posted it on the NBC website for all to see. His top six were: (1) Chelsea, (2) Man City, (3) Arsenal, (4) Man United, (5) Spurs and (6) Liverpool. These were exactly the positions at the end of the season.
Simon Gleave has collected and blogged about predictions for the last few seasons. In 2014-15 he found only 17 journalists who were brave enough to predict the entire season and publish it openly. Out of these 17, Prince-Wright was the only person to successfully predict the top 6. The most common error was to place Liverpool above Spurs or to overate Arsenal.
But Prince-Wright’s predictions weren’t perfect. The table below gives the difference between his predictions and the actual outcome.
The difference is zero for the top six, but is bigger (either positive or negative) further down in the table. He underrated Swansea and Southampton, and overrated Hull and Newcastle. The average distance between the prediction and the outcome is 2.3, implying that Prince-Wright was typically out by just over two positions.
To compare Prince-Wright to the other journalists and experts, below I plot each expert’s average prediction difference. The 17 dots are the 17 journalists. Joe Prince-Wright was second best and Mark Langdon was best.
The interesting thing is the solid line. This line shows what the average error would have been if we had simply written down the outcome of the 2013-14 season as the outcome of the 2014-15 season.
This is the most naïve prediction possible: teams will do just as well next season as they did the season before. But it was a good prediction. Using last season’s table was better than Prince-Wright and 15 of the other ‘experts’. Only Langdon outperformed last season’s prediction. The other experts did worse than it.
There are a couple of caveats. The change from 2013-14 to 2014-15 was unusually small. With the exceptions of a dip in form by the Mersyside teams, the late season collapse of Newcastle and rise of Leicester, most teams were surprisingly constant between these two seasons. So to be fair to the experts, we should rather compare them to the dotted line in the figure, which shows the typical performance of using last season to predict the next season.
When comparing to the dotted line, 5 experts are above it and 12 are below. This is not really evidence that they can beat using the last season as a prediction. If we go back another season to 2013-14 the experts do even worse. Only one of them beat the dotted line. There is little evidence here that experts can predict the outcome of a football season.
One interesting thing, however, is that Prince-Wright was best in 2013-14. He has a record of prediction. If he does well this season, we should start to be convinced of his predictive powers.
What are his predictions then? Well, in August Joe predicted a tight race between Chelsea and Manchester United for the title. It is certainly not too late for that to happen. But the results and the current odds are against him predicting the winner.
Simon Gleave has collected 92 predictions for the coming season, 39 are models, 28 are from fans, four are from bookmakers and the rests are from journalists, including Michael Cox, Nick Harris, Phil McNulty and Mark Langdon. I have to admit, I chickened out of making a prediction (other than to say it will be the same as last season) after I made this analysis. I don’t think its possible. Simon will present these predictions soon on this blog. It will be interesting to see if the expert’s predictions improve.
Scoreboard Journalism 