We were intrigued to figure out whether or not there is a relationship between the success of a team and the number of transfers they do during a window? This led us to create our hypothesis that, through the course of the 17/18 and 18/19 football seasons, the number of players purchased will have an impact on the success of a team’s next season.
We included four KPI’s (Key Performance Indicators) in our model which were absolute rank, absolute points, while relative rank and relative points compared to the previous season.
La Liga
In La Liga, on average, teams purchased about 6.76 players per season. The median number of players acquired across all teams is 7 players. The mode is the most frequent amount of players purchased, which is 1.
The standard deviation is 4.78, which means that there is a 66.6% chance that clubs will buy within the range of +/- 4.78 players from the mean value and a 95% chance that they will buy around +/- 9.56 players from the mean value.
The range of players purchased goes from 1 through 17 and there were 17 teams included in this data set.
OUTLIER
To determine whether there were any outliers, we ran an outlier test on the 17/18 points change values from the Spanish La Liga. The following table is calculated from the Spanish La Liga Change in Points values:
Giving us a Outlier test of [-23;17]
Evaluating the data, there were no teams in La Liga that achieved a 17 point gain or a 23 point loss over the two season comparison, meaning that no outliers were present.
CORRELATION
Last but not least, we ran a full correlation model in Excel on La Liga which confirmed our conclusion that there is no correlation between players purchased and change in points vs. the previous season.
However, this also gave us a more-reliable relationship, which is a weak negative correlation. What we found interesting is that there is a moderate positive correlation (0.63) between the number of new players in the 18/19 season and the final points in the same season. Our interpretation is that the teams with more points are more successful and therefore wealthier, leading them to have more money to spend for the following season, thus enabling them to buy more players.
Ligue 1
In Ligue 1, on average, teams purchased about 3.94 players per season, which is significantly lower vs. La Liga, indicating less transfer activity.
The median amount of players purchased across all teams is 3 players acquired. The mode is 2, while the standard deviation is 3.49.
The range of players purchased goes from 0 through 14 and there were 18 teams included in this data set.
OUTLIER
To determine whether there were any outliers, we ran an outlier test on the 17/18 points change values from the French Ligue 1. The following table is calculated from the French Ligue 1 Change in Points values:
Giving us an outlier test of [-37;29]
Evaluating the data, Monaco lost 44 points compared to the previous season, meaning that they are a clear outlier. On the other hand, Lille gained 37 points compared to the previous season which is above the upper bound of 29, making it an outlier as well.
CORRELATION
We repeated a full correlation model in Excel on Ligue, which again confirmed our conclusion that there is no correlation between players purchased and the change in points vs. the previous season. However, this also gave us a more-reliable relationship vs. team group analysis, resulting in a moderate negative correlation. We once more found that there is a weak positive correlation between the number of new players in the 18/19 season and the final points in the same season. Our interpretation is similar to the Spanish La Liga conclusion, where the wealthier teams have more money to buy new players.
Comparison La Liga Vs Ligue 1
Comparing La Liga and Ligue 1, we found that there is no correlation whatsoever between players purchased and change in points compared to the previous season. La Liga gave us a weak negative correlation, while Ligue 1 gave us a moderate negative correlation, disproving our hypothesis even further. However, both leagues exhibited a positive correlation between the number of new players in the 18/19 season and the final points in the same season. Our interpretation is correct that teams in both leagues with more points are more successful and therefore wealthier, leading them to have more money for the following season and, thus, enabling them to buy more players. The correlation R is very similar in both leagues, (-0.42 and -0.36 for Ligue 1 and La Liga respectively) which statistically speaking is very close.
Winter Transfer Window
Next, we came to the conclusion that the 2018-19 New Players Total variable is actually composed of players acquired within two distinctive transfer windows. Thus, the total new players can be broken into two transfer windows, the winter and summer ones, meaning that they can have different impacts on the success of a team. Furthermore, despite having combined all 5 leagues, each league could be exhibiting different behavior and impacting the outcome or significance of the model.
The r^2 here is 0.589 and this model tops all previous ones in terms of predictive power. The significance F here is very low and the F value is relatively high, indicating a statistically significant model. The newly-created independent variables of 2018-19 New Players Summer and 2018-19 New Players Winter are now both statistically significant, unlike when being added together, as in previous models, the combined 18 New Players Total variable becomes insignificant.
The 2018-19 New Players Summer is close to a 95% confidence level while 2018-19 New Players Winter is above the 95% confidence level. This model reveals that the coefficient of the 2018-19 New Players Winter variable is negative, indicating that with every player purchased during the winter transfer window, the team will, according to the model, lose 2.8 points. On the other hand, the variable of 2018-19 New Players Summer has a positive coefficient meaning that every player purchased during the summer transfer window will gain 0.81 points for the team.
However, the 2018-19 New Players Summer variable is on a 92% confidence level, implicating it is less accurate than its Winter Player counterpart.
Conclusion
In conclusion, throughout the first models, it appeared that 2018-19 New Players Total was a statistically insignificant variable, countering our stated hypothesis. However, throughout our model experimentation, we successfully broke up the New Players’ total variable into two sub- variables, 2018-19 New Players Winter and 2018-19 New Players Summer, which ended up yielding surprising results. The model demonstrated that players purchased during the summer transfer window would have a positive impact on the team’s performance (at a 92% confidence level) the following season, while players purchased during the winter transfer window, would negatively impact a team’s performance. In our opinion, and through various examples we witnessed, players purchased during the winter transfer window negatively impact their team due to several reasons, such as squad balance, moral changes, and motivation.
Sport Business Analytics 