Oxford's oldest student newspaper

Independent since 1920

Shots and Statistics

How far can we rely on statistics to tell the true story of football that so many will tell you are hidden from mathematical analysis?

After Chelsea dispatched Manchester United 4-0, Graeme Souness announced the death of stats in football. Despite losing so comprehensively, Manchester United had 56% of possession and crucially outshot their opponents 16-14. Souness couldn’t understand how on earth stats could be a useful metric to view a game which had such disparity between the score line and the simple match facts.

Souness was correct in one sense; comparing shots is an awful lense through which to view a game of football. This doesn’t, however, mean that statistics should be banished from the game: the wider football community needs to educate itself with better use of statistics. Instead of dogmatically listing possession, shots and shots on target, more sophisticated indicators of momentum and positioning offer us a way to understand nuances of the game and analyse critical moments in detail.

One method on the rise is expected goals metric. Instead of simply comparing shots, the quality of the chance is reflected. Weightings are given to each chance dependent upon a number of factors; distance from goal, type of assist, number of defenders.

If my memory of the game serves me correctly, a large number of United’s efforts were speculative attempts from Paul Pogba which had very little chance of producing a goal, and thus would have been given little weighting. Using historical data, we are able to roughly estimate the probability of any chance being converted, with cumulative chances being added throughout a game to leave us with an expected score line.

The map, courtesy of Michael Caley, reflects Chelsea dominance in the game (the larger the square, the larger the associated chance). Despite being outshot, the general quality of the chances they had were far greater than those of Manchester United’s, helping to explain Souness’ puzzlement.

The methodology is not perfect and is still being developed, but in terms of providing an idea of overall performance in a game, is far more useful than the stats we are currently provided with. Given the highly visual nature of “XG maps” it would come as a surprise if at some point down the line expected goals doesn’t become a regular feature of television coverage.

Expected goal data can also be viewed over the course of a season, once again providing us with a barometer of general performance as we look to potential unearth any lies of the league table. Compiling Michael Caley’s XG data from the first 9 game weeks we are left with the revised Premier League table below:

14875054_10211209789983481_966508823_n

There is a great deal that can be interpreted from our XG table. The season is still young and great unpredictability lies ahead, but with tools such as XG we can begin to better understand what is likely to pan out in the remaining 29 games. Next week, I’ll look to explore this, as the Premier League gets a quarter term report.

Graeme Souness was wrong—stats and football marry beautifully. You just need to use the right stats.

Check out our other content

Most Popular Articles