Possession

By Alex Rathke

How was the game? Which team dominated the game? Probably two of the most commonly asked questions by football fans. Reading between the lines, possession is often hinted at within this.  The statistics for possession are shown within 10-15 minutes of the game starting and tops the list on TV of stats at half and full-time. What is it, how is it calculated and does it mean anything? The first two questions can be answered by Opta – a sports data company based in London, England. They record possessions and define them as “the number of passes a team has in a game”. Now besides being extremely vague, how accurate is this for “measuring possession”? Not very or at all. This piece by Chance Analytics is well worth a read about measuring possession. Elsewhere, possession is used to describe one team’s dominance over another. While it has its benefits and can tell us which team was more dominant in holding onto the ball, it does not necessarily tell us much more than that. The last big question: Do teams with more possession generally win the game? Let’s have a look.

With Spain, Barcelona and most recently Manchester City playing that beautiful possession-based football, we may tend to think so. Better teams will naturally have more possession of the ball and look to use it to their advantage. Yet, the amount of possession a team has during a game does not necessarily guarantee a win. Why may that be? For example, if you continuously keep the ball but make no effort to move forward into the opposition half, possession of the ball is effectively useless. It is what you do with the possession that counts. A team that repeatedly penetrates the opposition half and creates chances = a higher chance of scoring. How does this apply to the PSL for the 2017/18 season?

The PSL

With data collected from Instat, we have been able to collect possession statistics for 212 out of 224 games (95%) success rate. It is better than nothing right? In the bar chart below, are each of the team’s average possession numbers. League average is coloured red, while blue represents the Top 8 and orange represents the bottom 8 teams. At first glance, we see the big guns at the top of the chart (Orlando Pirates and Mamelodi Sundowns) while Lamontville Golden Arrows are routed to the bottom of the table. Any surprises?

Personally, we don’t think so. For example, Pirates under coach Milutin ‘Micho’ Sredojevic played possession football before he took over last year. He has constantly through-out the season spoke about re-building and bringing stability to “The Buccaneers”. While a fourth trophy-less season later has come and gone we expect Pirates to have enough re-building done by the time August comes around again. Sredojevic has after all won trophies across the African continent.

As already mentioned, Lamontville Golden Arrows are rock bottom of the possession table. This by no means says that they are bad, (although they have gone through a rough patch this season themselves) quite the opposite in fact. The Backheel Boys have scored the joint 3rd highest number of goals this season (36) – Maritzburg United being the other team. Possession just is not their style so the opposite applies; counter-attacks.

Using Transfermarkt.com, this past season Arrows often played either a 4-2-3-1 or a 4-1-4-1 formation with fast attacks and width. You only need to look at the speed of Lerato Lamola, Knox Mutizwa and Kudakwashe Mahachi. On his second loan spell at Arrows, Mutizwa’s 7 goals and 4 assists may be all Arrows get out of the 24-year old Zimbabwean as his loan spell ends on June 30, 2018. Will he stay or will Wits recall his services? They could do with a player of his quality also.

Statistical Evaluation

It has been an interesting statistic to follow in the PSL this season. In 212 games in which possession data was collected and analysed, winning teams, 32% of the time had a higher possession statistic than their opponent. While this figure is low, there are some considerations we have to take into account. This values possession overall and not in specific areas of the pitch where the difference in winning/drawing/losing takes place. Without this information, we are taking the possession numbers at face value which would be a big mistake. Other factors to consider are league & team quality, tactics and even managerial preferences.

Diving even deeper into possession, the following was observed. Using a correlation test to measure points earned and amount of possession in the PSL, a perfect correlation value would return 1. A correlation test is a statistical measure that describes the association between random variables – in this case, possession and points. The result was r = 0.65, fairly high. So what does this mean? The high positive relationship means that as one value goes up, so does the other. This makes sense by understanding that generally, a relationship exists between the amount of possession you have and how many points you have in the league table. However, a gentle reminder is in order.  Correlation does not imply causation – the fact that if there is a correlation between two things, that it does not imply that one is necessarily the cause of the other. To note, we have also not accounted for squad depth, team quality, fitness levels, tactical situations and so forth.

Next up, we can use another statistical test (r2 – the coefficient of determination), which would indicate the proportion of points variance that is explained by the difference in possession alone. The r2 value here is 0.42. Again, please explain! Over the past season, 42% of the variance of the league table can be described by possession.

To put this into perspective, Ralph Honigstein –  a German Football journalist, examined the same correlation statistics using data from the German Bundesliga and English Premier League. He found that the r2 values were 0.50 (Premier League) and 0.65 (Bundesliga). This meaning that “50 and 65 percent of the variance in points can be explained by the factor of possession”. While there is a slight difference between the two European leagues and our own PSL, we must state the possible reasons for this difference.

1. This is a different league and the standard of two of the most elite leagues’ in the world cannot be compared.
2. One season of PSL possession data is smaller than 3 seasons of Bundesliga and Premier League data.

The ideal scenario would be to have confidence that rising possession statistics would equal rising points tallies. However, that is not always the case. Possession of the ball is just not every team’s strong point and teams should adjust accordingly. This reinforces the types of questions performance analysts should ask and repeatedly question with data. All that has changed is that information has become more detailed and finding the needle in the haystack may have gotten easier.