I had a recent conversation with David Butler in relation to his post on pundit predictions on English Premier League games. As we got talking about the predictions, David made a particular observation that very few 0-0 draws were predicted by the pundits that he looked at. We thought that this was interesting as behavioural economics could be used to explain this phenomenon. It could be argued that individuals have a tendency toward a ‘something will happen’ prediction bias (this is my own made up name – I think Dave has a more technical term for it!). That is, when you are asked to predict a football score in particular, it is more often the case that an individual will predict that something will happen rather than nothing will happen. The automatic or impulsive part of your brain (System 1) works against the reflective or controlled part of your brain (System 2) and comes up with predictions which are not rational given what has happened in the past. Perhaps people think that predicting that the status quo will hold is in some senses, not considered to be a prediction.
I pointed out to him that Mark Lawrenson also does Premier League predictions in his role as a pundit for the BBC so our conversation ended with my proposal to analyse his predictions against the actual outcomes and see whether there is evidence of a bias. As an expert we might expect Lawrenson to rely more on the reflective or controlled part of his brain, he knows that 0-0 draws occur more often than what a non-expert might think. Or maybe not. I used data from the current English Premier League season covering matches up to last weekend (7th March 2015, Total Games = 279). The following figure plots the distribution of actual scores against Lawrenson’s predictions. One can see that the distribution of actual scores is a lot more random that the distribution of predictions. Lawrenson has a particular tendency to pick 2-0, 2-1, 1-1 and 0-2 as final scores.
I have made some other calculations in the below table. Lawrenson also predicts more home wins that what has actually happened, conversely less away wins and less draws. His predictions have also never involved a game with more than 4 goals despite that fact that 26 of these games have occurred so far in the Premier League. Finally he has a tendency to over predict wins by small margins (2 goals or less). But what about the ‘something will happen’ prediction bias specifically? There have been 22 nil-nil draws so far in the Premier League and Lawrenson has (so far) predicted zero nil-nil draws. Obviously this is only one person/pundit but the fact that there is such a big difference between the actual and the predicted and the fact that Lawrenson has not even predicted one nil-nil draw is I think reasonably strong evidence to support the ‘something will happen’ prediction bias.