2016 saw possession become a dirty word in football. Leicester City showed that the Premier League could be won without the ball; an average of 44.8 per cent of possession, the lowest for any English champion in nine years, was enough for the Foxes to secure a sensational championship while Louis van Gaal’s Manchester United bored everyone to death with slow, methodical passing combinations that largely went nowhere.
In the Champions League, possession high-priest Pep Guardiola was knocked out by the transition brilliance of Diego Simeone’s Atletico Madrid, and at Euro 2016, Spain, the chief international team proponents of possession football, crashed out early in their second consecutive tournament in a row. Portugal sucker-punched France in the final despite being out-possessed (47 to 53 per cent), and RB Leipzig sporting director Ralf Rangnick claimed that Germany’s main problem in the semi-final defeat against the hosts in Marseille had been the enjoyment of “too much possession”, by which he meant a build-up involving too many touches.
Just when you thought that possession stats could be dismissed as irrelevant if not down-right misleading, one important figure made a surprising intervention recently.
In the wake of Man Utd’s 0-0 draw at Liverpool, José Mourinho re-appeared for an encore in the Anfield press room, insisting that his team had controlled the ball 42 per cent of the time, not 35 per cent, as the Premier League’s official number crunchers Opta had calculated. According to their data, United had never seen less of the ball in the Premier League since possession stats were first collated in 2003.
The explanation for the discrepancy between Utd’s and Opta’s stats most probably lies in the different methodologies applied. While some analysts measure the actual time a team controls the ball, Opta calculates possession as the percentage of overall passes. The number is informed by a team’s passing frequency and ability, as well as their ability to prevent the other team from passing: it can be seen as a metric for the movement of the ball.
That still begs the question why Mourinho should care, though. The Portuguese coach has in the past very happily played a clever reactive game, most notably at Internazionale, where training-ground-devised counter-attacking football yielded a Champions League trophy in 2010.
He was also well within his rights to point out that Man Utd “controlled the game for most of the time” despite Liverpool out-passing them almost two to one. But the 53-year-old was obviously irked by the low possession number, and for a good reason, too: while individual match possession stats tell us nothing about the two team’s relative efficiency, they remain a far more accurate gauge of a team’s ability to win matches in the long run than many realise.
A look at the last three full seasons in England and Germany shows as much. The top flights in both countries from 2013 to 2016 contained many different styles and special cases, including a Bayern side achieving absurd heights of possession, Leicester’s minimalist excellence in 2015/16, the various guises of Brendan Rodger’s Liverpool, teams like Hertha BSC and Köln who do remarkably well with very little possession and van Gaal’s Utd who did very little with a lot of the ball. But the general relationship between possession and success is far from random in both leagues.
As we can see on these linear regression graphics kindly prepared by Statsbomb’s Colin Trainor, utilising Opta’s official stats, clubs with higher average possession tend to amass more points per game in both leagues. There’s more variance in England, with some clubs’ points tally diverting dramatically from the line of best fit but overall still a high positive correlation (r = 0.70) between possession and points won in the Premier League. The correlation is even stronger in the Bundesliga (r = 0.80).
r = 1 would delineate a perfect correlation, with all clubs being on the same line, steadily, predictably picking points in an orderly fashion as their possession stats increase. But this is clearly not the case here. SV Darmstadt 98 (2015/16), for example, are an obvious outlier in the Bundesliga. They picked up 1.12 points per game last season with 37 per cent possession, an incredibly low number. No other team in either league had less than 40 per cent of the ball on average in three years, it simply doesn’t happen in elite football. (That might explain Mourinho’s discomfort with the Liverpool stats).
Champions Leicester were even more of a statistical quirk: their possession in 2015/16 was so low (44.8 per cent) that they overshot the predicted points per game line by more than a point.
If these “mistakes” in the model are possible, how reliable is it as predictive tool? We do know that other factors that we haven’t calculated - squad quality, depth, fitness levels, etc - must have a bearing on points, independent of possession. But can we quantify the importance of possession itself? We can.
r2, the coefficient of determination, denotes the proportion of points variance that can be explained solely by the difference in possession. Here, the value for the Premier League is 0,50 - it would be 0.58 without Leicester’s freakish win last season - and 0.65 for the more linear Bundesliga, where points and possession correlate more closely.
50 and 65 per cent of the variance in points can be explained by the factor of possession. Think about that for a moment. 50/65 per cent of the variation in points by any two teams that we know absolutely nothing about - they might be defensive or attacking, good or poor - can be explained by possession, and possession alone. It’s a hugely important factor, allowing us to predict with confidence that rising possession stats go hand in hand with a rising points tally.
Simply passing the ball around more than your opponent will not guarantee you points, of course. It would also be silly to advise a poor team to aim for high possession, as they wouldn’t be able to do it. But what the numbers do show very clearly, is that successful teams also happen to be better and more frequent passers, and vice versa. And that shouldn’t really come as a surprise.
Possession numbers imply high technical ability, good movement and the ability to stop the opposition from playing. For those on the other side of the equation, it’s of course still possible to succeed with fewer, better passes, but it’s also statistically less likely. To win points the underdog way, consistently without the ball, is much harder than with it.
Leicester, Portugal and Atletico might have given possession a bad reputation but it would be wrong to conflate their against-the-odds success with a new trend. These teams are outliers, exceptions to the rule.
Football remains a game that rewards those who play it.