Skip to content

lnoguera17/Conmebol-Analytics

Repository files navigation

Soccer Data Analysis of South American Soccer Players

Insipiration and Future Work

This project has been created solely for enjoyment and learning purposes, however, we believe that similar approaches can be performed at scale, capturing strong signals of a player's season, latest games played information and multiple data points beyond a soccer player's performance to automatically and effectively identify attractive buying opportunities during the transfer market window.

Web Scraping

Used the Rvest package to scrape SoFifa's website and collected data from more than 1,000 soccer players from South America. Data collected includes, overall player's rating, player's height, market value, best position in the field, ball control, dribbling, nationality and more.

Exploration Data Analysis

Wrangled, cleaned and visualized the scraped data using the Tidyverse ecosystem to generate and communicate quick insights.

Statistical Modeling

Fitted a base linear model to compare more poweful ones, pre-processed, resampled, fitted and evaluated linear models after multiple feature trasnformations and achieved 0.88 Adjusted R-Squared Error and ~EU 4M RMSE with a tuned Random Forest model. We used the Tidymodels ecosystem to understand drivers of the player's market value and provide an objective value estimation.

Dashboard and Model Deployment

Created a dashboard using Shinydashboards to provide users the ability to set new parameters for a new player and automatically estimate his market value. Also, the shinydashboard offer dynamic objects to visualize and provide information about the soccer players in the dataset. Access the Shinyapp here

Collaboration Tools

To keep track of the project scope, progress, stories and organization we used Notion, check out our notion page, here. This proved to be an amazing all-in-one workspace that allowed us, the contributors, to stay focused an organized during the whole project. We also switched our project conversations, git and github activity to a Slack channel Slack, this method we think, served as a great tool to keep up with new remote branches created and commits added.

Want to collaborate?

If you are looking to improve the statistical model or have creative ideas to incorporate in the dashboard, reach out to any of the contributors. Looking forward to connect and work together!

Project Contributors:

Luis Noguera

Nicolas Kaswalder

To all Barcelona fans, specially Nico Kaswalder

Barca 8-2

About

End to end data science self-learning project. This work focuses on R, tidyverse and tidymodels ecosystem for data wrangling and modeling.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages