FIFA 19 DS & ML Applications
Authors: Ekrem BAYAR, Alperen BALIK
Working with Shiny for 1+ years
Abstract: I first designed the dashboard to compare teams and player stats inside the Bundesliga, La Liga and Premier League. Newest additions and improvements made it possible for better visuals, new leagues, in depth ad-hoc analysis and EDAs to enhance the data science experience in sports analytics.
Full Description: There are 8 leagues in the dashboard such as Bundesliga, Eredivisie, La Liga, Liga Nos, Ligue 1, Premier League, Serie A and Süper Lig.
Leagues page contains descriptive statistics, visuals and comparisions about each leagues. Also it shows the best 11 players in terms of tactics.
Teams page gives some useful insights from all of teams.
• Summary tab shows us to learn talent distribution based on variables and position classes.
• Value tab visualize value of the team.
• Best Players tab helps to find the suitable tactics for the team.
• Stats tab containts top and bottom features and players in the team. Also there is a heatmap to compare players.
• Set Piece Goal tab tells us to find relevant players for free kicks and penalties.
• BMI tab reports and visualize BMI (Body Mass Index).
Players page is one of the important page in this dashboard. It helps to compare two players. There are some descriptive statistics and visuals. Also it finds similar players using distance measurements.
• Radar graph used some features to compare two players.
• Bar chart shows which feature is better between two players.
• Line chart shows players' value by years.
• Similarity tab discovers similar players using distance measurements. There are many different methods such as Eucledian, Maximum, Manhattan, Canberra, Minkowski, Pearson, Spearman and Kendall. Max three leagues can be entered for analysis.
Another important page is Scout page. Scouting is very important for Clubs. That's why, I tried to use some statistical methods to discover talented players. I also designed a surprise page with gamification. You can create a team and also transfer and sell players.
• There is a player database. You will be able to find players using the database.
• Best players tab shows top ten players from each position class.
• Player Stats tab contains SWOT Analysis, Player Data, Visuals, Hierarchical Clustering and Linear Discriminant Analysis (LDA). In order to find the possible position of the player LDA algorithm was used.
• Player Clustering tab enabes to apply K-NN algorithm. K-NN works for each position class and finds the clusters. I tried different k parametres such as 2,3,4 and 5. I chose optimal k value for clustering and I found k=4 as optimal k value. After the analysis, there are visuals, K-NN results and player clusters.
• PCA tab contains a multivariate statistical analysis. PCA means Principle Component Analysis. PCA gives a score to compare the observations. I used to compare players and teams each position class. I used to compare players and teams each position class with PCA Score. Bar graphs show us the best players by using PCA Score and Overall variable.
• Correlation tab helps to find relationship two variables. I did correlation tests like Pearson, Spearman and Kendall methods and ploted the results. Also I added a hypothesis test. Unpaired Two-Samples Wilcoxon Test investigates whether there is a significant difference between the two groups.
• I wanted Player Searching tab to be a game. There are some inputs to search players. You can see the players depend on inputs. You can create a team, transfer and sell players with this tab. You also have a budget, you should spend your honey carefully.
Category: Sports
Keywords: FIFA, Sports Analytics, Data Science, Football, Shiny Dashboard, EDA, Data Visualization, Video Games, Sports
Shiny app: https://ekrem-bayar.shinyapps.io/FifaDash/
Repo: GitHub - EkremBayar/FifaDash: FIFA 19 DS & ML Applications - Shiny Contest 2020
RStudio Cloud: Posit Cloud
Datasets: Kaggle
https://www.kaggle.com/karangadiya/fifa19
FIFA 20 complete player dataset | Kaggle
If you want to see old versions of the app, you can find all on my kaggle post.
Thumbnail:
Full image:
NOTE: Due to some problems on url, the player photos and the club photos have removed. 
You can see the broken link below. This link belongs to L. Messi.
https://sofifa.com/players/4/19/158023.png
It is going to be fixed as soon as possible. All of the broken links will change with unknown player photo. (2020-04-13)
