¡HI THERE! 👋🏼
FIFA 21 is a football simulation video game developed by EA Vancouver as part of Electronic Arts’ FIFA series. As a football fan since childhood, I am excited to perform Data Cleaning, Data Wrangling, EDA and Modelling using FIFA 21 player dataset from Kaggle.
The dataset contains information of all 17125 players from FIFA 21 edition. There are 106 attributes including Age, Nationality, Overall, Potential, Club, Value, Wage, Preferred Foot, International Reputation, Weak Foot, Skill Moves, Work Rate, Position, Jersey Number, Joined, Loaned From, Contract Valid Until, Height, Weight among others.
Perform an end-to-end analysis applying statistical or machine learning techniques and present the results.
Avoid Inadecuate pricing: A club might may spend too much for a player that is not worth or sell a promisiong at low price.
Providing insights: Assist player signings by filtering datapoints to provide useful information for scouts our staff dealing with buying/selling players.
Make accurate predictions: Predict market value of fooball players. Clubs can have the opportunity to assign a fair market price for players they want to buy and avoid paying too much for players and use wisely their budget.
- Data Cleaning: Handling missing values, removing duplicates, correcting data types, renaming columns, handling inconsistencies in data..
- Data Wrangling: Filtering rows based on conditions, sorting data, reshaping data, grouping data based on categories..
- Business Case: Let's imagine that one of the top teams in the world is interested in signing a player for the next season and has to make a decision soon. We will identify the best options taking as a sample the following questions:
- Who are the 10 best players according to their statistics and performance?
- Who are the best players in every position in the game?
- Who are the 5 players under 25 years old with the highest potential?
- Players with the highest release clauses and their Ratings.
- Which are the market values of Top 10 players?
- EDA: computing statistics, cratin visualization to explore data, exploring distribution, identifying correlations, detecting patterns and trends in data.
- Predicting Market Value: splitting data into Features, train/Test Split data, transform data for training a Linear Regression model, train and test models using normalized and standardized numerical data.
- Summarize results/conclusions.
🔹 Strong correlation between wage, release clause and their market value.
🔹 Players with a high release clause and wage is likely to be more valuable in the transfer market because it indicates that they are in demand and their current club values them highly. Release clause and wage could also be influenced by a player's performance and potential.
🔹 The determination coefficient hardly varies by eliminating the categorical data from the model. When I eliminate the variable that has a high correlation with our predictor variable, release clause, r2 falls below 0. In the next analysis round I’ll test how it responds with a model focused only on ‘n’ numerical variables.
🔹Presence of many extreme outliers in the distribution of certain characteristics could have skewed the analysis and made it more difficult to draw meaningful conclusions.
🔹 Additionally, not taking into consideration account the player's position may have also impacted the results, as certain characteristics may be more important for certain positions on the field.
The model and errors could be improved by trying other statistical methods such as robust regression or random forest.
Overall, as a football lover this project was a great opportunity for me to apply my data analysis skills in a practical setting. I learned a lot about FIFA 21 and the factors that contribute to a player's value.
I ❤️ FIFA! 🎮
During the project I used Python 3.8 as programming language and the following libraries:
- Pandas
- NumPy
- Matplotlib
- Seaborn
- Sklearn
- SciPy
- Math
Emilia L. © 2023