Project1_MoneyBall_FIFA

Objectiv

Performing an end-to-end analysis putting into practice what I have learned so far. Get a model to predict the value of a player.

Subquestions:

1. Order players by Overall Rating (OVA).
2. Which country has in average the best players (by OVA)?
3. Are the best 10 players by OVA more or less agressive than the average?
4. Where are most of the players from?

Dataset

In this project, I used the provided fifa21_male2.csv dataset.

Here details about the dataset can be found here as well.

This data set includes:

1. EA Sports FIFA 19 Game data:

   Player Name	Club of the Player
   League	Position
   Pace	Shooting
   Passing	Dribbling
   Defending	Physical

2. Transfermarkt extra info by player:

   Date of Birth	Nationality
   Height	Foot
   Day Joined the current club	Day of Contract End
   Market Value of the Player

3. Instagram and Facebook data by player:

   • Number of followers on Instagram
   • Number of likes on Facebook of the club in which the player has a contract

4. ESPN FC data from the past 5 years performance of each player

   • GS: Games Started
   • SB: Games Substituted
   • G: Goals Scored
   • A: Assists
   • SH: Shots
   • SG: Shots on Goal
   • FC: Fouls Committed
   • FS: Fouls Suffered
   • YC: Yellow Cards
   • RC: Red Cards

Workflow

First, I answerd my suquestions and than started to clean the data. In this process I droped some unnessecary columns for some different reasons. For example, I did't needed the names and pictures/logos for my regression model or some data wad shared multiple times in the dataframe. After some general cleaning the data, I check the correlation of some caolumns and droped some that were unnessecary. After applying a boxcox transformation, I checked the p-values and droped some more columns which had an high p-value. Furthermore, I removed the outliers for all the columns.

To get the best result I compared the boxcox transformation to a normalization. The boxcox was best with an R2 of 0.9551 after removing columns with an high p-value

Answer of the subquestions

1. See notebook.
2. In average, the player in Tanzania, Syria and Mozambique have the best overall rating.
3. The agression of the top 10 players is a little bit lower than the average.
4. Most of the player are from Englnd (1707), Germany (1154) and (1121).

Libraries

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import seaborn as sns

import scipy.stats as stats

import os

import statsmodels.api as sm