Using data and machine learning to analyze football players.
Contributors:
- Baturalp Yalcin
- Thomas Mecattaf
- Mohnish Chakravarti
Description: All of us play FIFA career mode on our xbox almost every day. In this project we analyze in-game FIFA ratings and real-world performance for EPL players from the 2015/16, 2016/17, 2017/18, 2018/19 seasons using basic machine learning techniques such as linear regression, random forests and neural networks.
There are 3 notebooks (one for scraping, one for basic data analysis and one for machine learning), and 3 HTML files that explain all of these notebooks and our project in more detail
Some motivations for doing this project are:
-
Can we make sense of player attributes and ratings in FIFA? EA Sports does not release any information as to how players are ranked and assigned attribute values. This is turn can be used by FIFA players to purchase better players for each position their teams, since we would know which attribute(s) is the most important for each position. (I only used to look at sprint speed for each position.. now I know that I should not!)
-
Can teams use similar techniques to assign rankings and ratings to players in real-life and determine who would best fit into the team's overall strategy? (cough Alexis Sanchez and Courtois cough) This is not only useful for teams looking to spend big money on the right players, but also for youth teams looking to better evaluate players and promote them to the first team at the right time, as well as cash strapped teams that cannot go all-out to purchase the best players for each position
-
Betting - you can build a probabilistic model for matches based off the information for each player (and then for each team) to predict scores for matches (no, we haven't bet using our work)
Some libraries that you should install: BeautifulSoup, selenium, CSV, pandas, numpy, scipy, seaborn, matplotlib, os, time
Everything is in ipython notebooks so you should not have any issues running anything, but if you need the datasets (so that you can skip the time-consuming web scraping bit) or have any questions, feel free to email me at baturalpyalcinn [at] gmail [dot] com