“Possession of the ball is the key to winning” - J. Laing Burns, Jr.
Many footballing philosophies live and die by the idea of keeping possession of the ball. It has been drastically implemented in the Pep Guardiola system, which endeavours to methodically progress an attack through possession.
A recent study conducted as a project for the International Baccalaureate program (as part of a mathematics internal assessment) is showing that winning football matches and having more possession are related by more than chance.
The study, conducted by an IB student, looked at the last 100 English Premier League matches that had differing possession statistics between the two battling sides (meaning the possession was not 50/50).
Using live.premierleague.com to gather the data, it was discovered that the team with more possession (above 50 percent) won 44 times, lost 34 times, and tied 22 times. This means that the team having more of the ball only won 44 percent of the time in the last 100 EPL matches. This would seem to indicate that possession is not related to winning football matches. Right?
Well, not mathematically. In this investigation, a chi-square test was used to determine whether winning and possession were independent, or if the they were related by more than chance. For those of you who do not know what a chi-square is, it’s a math test used to see whether two things are related or not. Here, the test was between winning and not winning.
Article continues below
To test this, the chi-square was divided between low and high percentage of possession and winning and not winning.
Possession - low: 0-49.99%, high: 50.01-100%
Wins - win: wins, no win: losses and draws
Then, the observed chi-square matrix was filled out. Essentially, this is just putting all the data into a form that can be used for the test. The team with the least possession won 34 times, and did not win 66 times. In contrast, the team with more possession won 44 times, and lost 56 times. That was inputted into the chi-square matrix.
With that information, an expected chi-square matrix can be filled out. This shows what you would expect to find from the amount of gathered data. This can be completed by adding the columns and rows of the matrix and the total amount of numbers in it.
Then, in an individual box, the row that box is in is multiplied by the column that box is in, and then the product is divided by the total amount. This leads to the expected data for that box (in this case, it would lead to what you would expect to see from a team who had less possession to win or not win, and the same for the team who saw more of the ball).
Afterwards, using a calculator, p was determined. In conjunction with a critical value (determined using degrees of freedom which in this experiment is 1), the relationship, or lack thereof, can be determined. In this case, p was less than the critical value, which leads us to reject the original hypothesis and accept the alternative one.
The original hypothesis is always the same. The two subjects are independent of each other (possession and winning a match are not related). The alternative is that the two are related by more than chance (possession and winning a match are related by more than chance). The investigation proved the alternative hypothesis: possession and winning are related by more than chance.
In conclusion, the investigation simply showed that possession and winning are related. Although the team with more possession only won 44 percent of the time, that was significantly more than the 34 percent win ratio teams with less possession experienced.
With that in mind, personally, if I were a football manager (which I guess I am thanks Sports Interactive, Sega, and the creators of Football Manager 2015) I would set my team out to keep the ball.
However, this study did not prove what the exact relationship between winning and possession was, only that there was a relationship.
Also, this study did not test where the majority of the possession for the dominant team was on the pitch (for both winning and losing teams), whether or not teams were home or away, or the number of shots each team had. That would need to be studied in a further investigation to see how those factors may have affected the outcome of each match.