Case Study: FIFA-MoneyBall
Brief overview of analytical processes
Objective:
Build a regression model based on the train data set to predict the OVA score for footballers in test data.
What we did:
• We started by studying the data and noticed that different types of players had different abilities and playing styles. For example, a goalkeeper has different skills than a midfielder or a defender.
• Created 4 regression models for 4 different categories: Goalkeeper, Defence, Midfield and Attack
• Applied this data to the test data (seperated this data also into 4 postiions: Goalkeeper, Defence, Midfield and Attack)
• Got the 3 footballers with the highest OVA and under the budget
The mean absolute error (MAE) for our regression models – a measure of how accurate our predictions were – were as follows:
Goalkeeper
0.2745587887799759
Defence
1.00683400767267
Midfield
1.3396425816041848
Attack
0.4362952456631845
We chose players from the following models: Goalkeeper, Defence, and Attack
Players that we chose:
M. Neuer predicted OVA = 88, value = €29M
L. Suárez predicted OVA = 87, value = €31.5M
P. Lahm predicted OVA = 87, value = €29.5M
We had €100 million in the beginning. After selecting the three players, we were left with €10 million, successfully managing to stay within our budget while selecting top-rated players.