Not All Possession is Created Equal: Measuring Possession with Marcelo Bielsa
In this piece, Jon Mackenzie explores the notion of possession in football, how we measure it, and what it tells us about Marcelo Bielsa’s Leeds United…
After two full seasons of Marcelo Bielsa, a lot is made of Leeds United’s possession statistics.
In 2018/19, Leeds picked up the highest possession percentage in the Championship — 64.6% which put them well ahead of second-placed Brentford’s 58.6%. The following season, Fulham gave Leeds a bit more of a run for their money, putting up a credible 61.5% to Leeds’ 64.4% (although perhaps the interminable passing between centre backs was not worth the effort in the end?).
Newly-arrived in the Premier League, Bielsa’s Leeds have been brought back to earth with a bump. A small sample size it may be, but 54.5% is way off the heady heights of the Championship.
Of course, there are caveats: Leeds basically broke even with Liverpool possession-wise and pulled off a Brexit vote against Manchester City (52–48). However you look at it, though, despite the change of calibre in opponents, Leeds are still looking to possess the ball in the Premier League.
For the Leeds fan base, this is seen as something as a badge of honour: “Look at us! Going toe-to-toe with the biggest teams in the world per ball retention!” This is fair enough. Anyone who watched Leeds in their games against Liverpool and Manchester City will have been impressed by the parity. But are possession stats in any sense meaningful when it comes to measuring a team’s performances?
Manchester City, Liverpool and Chelsea are the top three sides when it comes to possession metrics this season. But the three teams at the top of the table as things stand are Everton, Aston Villa and Leicester, of which only Leicester are in the top five possessors in the league so far.
Obviously, sample size, yadda yadda. But even last season, although there is a correlation between league position and possession figures, it is not a strict correlation. Manchester City dominated possession stats (66.9%) and came second. Newcastle propped up the league (38.6%) and finished thirteenth overall. Norwich, who would eventually go down, put up a final possession percentage of 49.3: good enough to put them in the top half per the metric.
So why do we keep going on about possession?
What is Possession?
It might surprise you to learn that when you see a possession statistic as a percentage at the end of a game, it isn’t based on the amount of time each team spent on the ball.
In the past, data loggers used to use a “chess clock” approach to measuring percentage which required flicking a two-way button from side to side depending on who had possession of the ball. Understandably, this was fairly high maintenance so, in the end, a new approach was formulated.
Here’s how Opta describe the process now:
“Opta now record possession in a football match by means of an automated calculation based on the number of passes that a team has in a game.
We have two analysts, each monitoring one of the teams and they log each event in a game, totalling between 1600 and 2000 events per match. Each of these events has a timecode plus an x,y coordinate and the collection system is rigorously monitored by our team of checkers.
During the game, the passes for each team are totalled up and then each team’s total is divided by the game total to produce a percentage figure which shows the percentage of the game that each team has accrued in possession of the ball.”
For Opta, then, “ball possession” means the ratio of completed passes and is not a measure of time, although they claim that the two are very closely related.
Now clearly, the number of passes you make as a team will largely translate into how much of the ball you have which, in turn, will usually translate into something like in-game dominance. But as a metric, that isn’t telling us much at all about the flow of the game.
In fact, the most interesting thing about possession is that it is a binary phenomenon: that is, if one team has possession, the other team does not (and vice-versa).
This means possession is “shared” and “shared” in such a way that alternating chains of possession will be built during the course of games. If one team has possession of the ball 10 times during a period of the game, the other team will have had to have had possession 9 times or 10 times or 11 times.
Why is this interesting? Well, because it gives us another metric by which to approach the concept of possession so as to determine more about it.
Units of Possession
Imagine two teams are playing. Both of them usually dominate possession. In this instance, it is impossible for that to occur because their possession percentages have to sum up to 100.
At the end of the game, one team has won comfortably. But strangely, the possession stats show up as 50–50. Why should one team have prospered more than the other in this situation?
Looking at the raw possession value which, as we saw, is based on the number of passes made, we are probably none the wiser. But by considering “units” of possession now and summing them up, we might be able to get a clearer sense of what’s going on.
Let’s return to our two hypothetical teams again. Imagine we do a bit of digging and look at the units of possession for their previous games this season.
When it comes to Team A (the team who won comfortably), we find that, on average, their games saw 100 turnovers of possession. This means that, when they played, they saw 50 units of possession per game.
Team B, however, only averaged 50 turnovers of possession per game or 25 units of possession for each game.
Returning to the match played between these two teams, we now have a better approach to possession by which to analyse the game. As it turns out, the units of possession for the game was 86 or 43 units of possession per team.
This figure puts the game much closer to the conditions that Team A usually enjoy and so could explain why they came out so comfortably on top in the match.
A Closer Look at Units of Possession
The Athletic’s Tom Worville looked at this idea of units of possession in a piece at the end of last season. Here’s the data visualisation he drew up for the Premier League:
As you can see, the team with the most units of possession are Southampton. On average, they saw over 100 units of possession during their games.
Given that the amount of time in a game is roughly equal (although much lower than the allotted 90 minutes when you take account of how much time the ball is out of play) the more units of possession there are, the shorter those units of possession must be on average.
For Southampton, then, this tells us something about how they play. Their units of possession are shorter but, importantly, so are their opponents. This tallies with the fact that Southampton are managed by Ralph Hasenhuttl, a famous proponent of gegenpressing: a direct, counterpressing style of football in which turnovers of possession are treated as opportunities for attack.
At the bottom of the table, we see Manchester City, whose unit of possessions per game rolls in at an average of 85.
What does this mean? Well, we know that Manchester City are a high possession team who use a high press as a defensive measure to win back possession once they lose it. This adds up to what the data shows: a high possession team will see longer stretches of possession which, in turn, given match length is roughly equal, means that the units of possession will be lower.
At the end of the piece, Tom Worville is keen to stress that we shouldn’t read anything into the ordering of these teams by units of possession: “This is a purely stylistic measure — there’s little to no correlation between the ordering of the teams here and their position in the league table — but that doesn’t stop this from impacting performance or being a worthy measure to consider.”
So how might this operate as a worthy measure of what is happening at Leeds United?
Leeds United and the Unit of Possession
After reading Tom’s Athletic piece on possession, I asked him if he had a similar graphic for the Championship last season and he was kind enough to oblige:
What do we learn about Leeds? Well, perhaps surprisingly given their congruence to Manchester City in so many other ways, Leeds find themselves towards the top of the units of possession table with well over 100 possessions per game.
Does this tell us anything? Primarily, we know that, while Leeds and Manchester City had similar possession shares — ~64% — this translated into over 15 more possession phases for Leeds in a game than Manchester City.
Relatively speaking, this means that Leeds don’t control possession in quite the same way as Manchester City — they have shorter possession sequences when they are on the ball but, although they do give the ball away, they are also very good at winning the ball back so as to maintain a possessional advantage over their opponent. (You could look at something like Passes Per Defensive Action as a metric to confirm this.)
In short, this suggests that Leeds are going to be more transitional than Manchester City, something that bore out in the match between the two teams that took place before the international break. It wasn’t until around the 75th minute that Pep Guardiola accepted that, in a transitional game, his team were playing to Leeds’ strength. Bringing on Fernandinho and slowing the game down brought the game back onto City’s terms.
This season, Leeds have continued to dominate one possession metric: the unit of possessions table. On average, Leeds have seen around 102 units of possession in their games in the Premier League so far. This suggests that Marcelo Bielsa intends to continue playing the way he has been playing in the Championship.
But whilst we may have been tempted to equate that possession with careful control in the mould of Manchester City, the underlying data suggests a slightly different playstyle: one that focuses on transition and then quick recovery on the turnover.
It’s easy to make an assumption about a football team’s playstyle off the basis of a simple possession stat. However, the concept of possession remains largely under-explored as a source of interesting data.
In this article, we’ve simple talked about raw units of possession but each of those units will be different lengths. By taking into account possession length as a unit of time, then we could start to tap into a whole aspect of football tactics that remains underdeveloped.
For example, how does game state impact on possession units? Could we trace a game through the changing length of possession units? How might a club prepare for opponents by working on breaking down their possession unit (or, alternatively, stretching it out) to their advantage?
Hopefully, this article will have opened your eyes as to how much more we still have to learn about possession in football. Watching Leeds United under Marcelo Bielsa is always an eye-opening experience, though, so no doubt a more expansive idea of possession will allow us to understand his tactics even more fully as time goes by.