Creating-a-Pass-Success-Probability-Metric-with-Decision-Trees

In this project, we used tree-based models to compute a pass success probability metric. Tree-based models tend to be good predictors given that they normally allow for both regression and classification, and they usually stand out for being very interpretable.

In terms of applicability to the sport, the pass success probability allowed us to separate those easy-to-perform passes from more difficult ones. This yielded information on player behaviour, since we saw that some players tended to perform riskier passes, whereas others preferred to play it safely. But it also showed how good a player was as a passer if they tended to succeed when performing difficult passes.

The dataset we used contained information about all the passes performed during a season of a specific competition.

This project was provided to me as a graded exercise as part of the Sports Analytics Postgraduate I was part of this year (2022-2023). It was designed by FC Barcelona experts working in the Data Science team which are also teachers in this postgraduate that was created by FC Barcelona and UPC School.