Friday 18 October 2019

Hope springs eternal but the statistics reveal truth

Tom Brady leads the New England Patriots against the Chicago Bears tonight with, according to, a 60 per cent chance of victory. Photo: Getty
Tom Brady leads the New England Patriots against the Chicago Bears tonight with, according to, a 60 per cent chance of victory. Photo: Getty

Greg Wood

How well do you know your football team? And how do you think they will get on next weekend? Because my bet would be that however well you think you know your team, no matter how dispassionately you feel you can assess their chances after season upon season of occasional highs and abundant lows, there is a database in the United States that knows a great deal more about your club's performance than you do.

In fact, there are quite a few databases which contain more wisdom about the nitty-gritty of how and why football teams succeed or fail than any human could ever hope to learn. But the vast resource of sporting information and analysis on the website is of particular interest for two main reasons. The first is that they give it all away for free. The second is that it was founded by the statistician Nate Silver, who can fairly claim to have written the book when it comes to using statistics, data and probability to predict what will happen in sporting events, elections and much more besides.

The book in question - The Signal And The Noise: Why So Many Predictions Fail, But Some Don't - was published in 2012, a few weeks before Silver improved on a 49-out-of-50 record predicting how the states would vote in the 2008 presidential election with a perfect 50-out-of-50 next time around. Politics and election modelling is an important part of FiveThirtyEight's output but Silver set out in life as a baseball analyst and predicting the outcome of sporting events is another vital part of the mix.

Across a year FiveThirtyEight runs a statistical, data-driven rule over more than a thousand different teams, competing in tens of thousands of matches, and assigns each one a chance of success, expressed as a percentage. Its figures suggest that Everton have a 44 per cent chance of beating Crystal Palace today, for instance, while the New England Patriots' trip to Chicago tonight is assessed as 60-40 in the visitors' favour.

It is an approach that sits well with American sports in which draws tend to be seen as one step up from communism and terminated with extreme prejudice wherever possible. Football's three-way proposition is a little more complicated but since January 2017 FiveThirtyEight has risen to the challenge, initially with forecasts for six domestic leagues and now for no fewer than 34 domestic leagues, on four continents, as well as both the Champions and Europa Leagues.

And it has done so to impressive effect. Imagine that in a particular game, the prediction is a 50 per cent chance of a home win, a 30 per cent chance of a draw and a 20 per cent chance of an away win; that means an away win is unlikely in any one match but across five such games you would expect it to happen once. In FiveThirtyEight's predictions for the range of chances between 20 per cent and 60 per cent, where the majority of outcomes in football matches lie, you can see this principle at work. In this season's Premier League, for instance, 97 possible outcomes have been forecast as being in the range between 21 per cent and 30 per cent. These have occurred on 26 per cent of occasions, smack in the middle, and much the same goes for the ranges from 31 per cent to 40 per cent (current strike rate: 32 per cent) and 41-50 (current rate: 43 per cent).

It is not quite a straight line but the season still has a long way to go and any kinks are more likely to straighten than increase as time goes on, and as the amount of data in the model increases.

Nor is it simply a case of feeding numbers in at one end and churning out a percentage at the other. The model is subject to regular tweaks and additions, a recent example being a new assessment of the significance of every match in terms of the effect it could have on the table. Will one or both teams be all out for the win, or might one of them settle for a draw? These additions all feed into the evaluation process, and their effect can be measured as well. In this case the site says: "The improvement . . . in our match forecasts when incorporating match importance is about one-third the size of the improvement we saw when we added expected-goals metrics."

Step-by-step, data analysis and statistical forecasting is getting progressively better at pinning down what can be expected from a match. And while this might seem to be of significant interest mainly to those who want to bet on the outcome, it has lessons, too, for fans of any team that have their ups and downs. Which is, of course, all of them.

What the number-crunching on a site such as FiveThirtyEight offers is an objective, unbiased assessment of a team's chance. For most teams in most football matches, the verdict is that their chance of a win is less than 50 per cent. Most fans, however, are optimists. It pretty much comes with the territory. No one goes to a game hoping to see their team lose and a majority - in all likelihood, an overwhelming majority - would guess that their chance of a win in any given match is at least 10 per cent higher than the forecast suggests.

So spare a thought for the manager, who must live and work in the lonely, unforgiving space between where their team actually are and where most of the fans think they should be.

In the Premier League at least half the managers set out with the overriding goal of reaching 40 points for survival but there will be only a handful of games, at best, in which their chance of a win is better than 50 per cent. So it should not surprise or irritate anyone if they take a pragmatic view when it comes to balancing attack and defence. Every team in the division is handed 38 points for nothing over the course of the campaign, at a rate of one every time they kick off a match. How many of those points they successfully defend can be the difference between staying up and going down.

Yet fans are irritated, often intensely so, if their team are not throwing everything at their opponents from the first whistle. That frustration can also feed through to the team and the board, in an age when three apparently "poor" results - in matches that the team were unlikely to win in the first place - can be enough to get a manager the sack.

Statistical forecasts such as those on FiveThirtyEight highlight the abiding difference between what sports fans want to see and what they should expect to see. It is only human to forget all about the expectation gap once the game is under way but also only fair to acknowledge it when the time comes to reflect - and recriminate.


Sunday Indo Sport

The Left Wing: Welcome to Irish rugby's biggest week - is an upset on the cards?

Editor's Choice

Also in Sport