Project · 2020

Like-for-like player clustering

A small tool for finding soccer player replacements by clustering on skill attributes — mostly an excuse to play with FIFA 19 data.

Stack Python · scikit-learn · Pandas · Seaborn

Role Coursework, then an evening project


If a team loses a player, who’s the closest available replacement on the transfer market? This notebook answers that by clustering FIFA 19 players on their skill attributes and pulling the nearest neighbors within a cluster.

Along the way I used the same dataset to compare Naive Bayes, logistic regression, and SVM on a position-prediction task — a decent sandbox for intuition on when each model wins.

← all projects