Create

Design Project Gallery

Project Search
Filter projects by keyword, program, course, or submission year.

Search Fields

Clear all filters

Soccer Expected Possession Value

Project Description:

Soccer decision-making is complex and dynamic. This EPV tool aims to quantify the best move in soccer at any moment, making decisions more intuitive. Our main motivation was how exactly we wanted to quantify this best move and how to relay that information to the user. We developed a computer vision pipeline that detects players, estimates their location, and integrates this into an Expected Possession Value (EPV) model to recommend optimal actions for players in a given game state.

Project Photo:

A field showcasing players as circles and arrows indicating the probability of a successful pass to them if they are on the ball holder's team.

The probability of a successful pass to other players on the ball holder’s team.

Project Poster

Open full size poster in new tab (PDF)

Project Poster Summary:

Summarized findings of a year-long project where a model was trained to predict the pass probabilities and the best move to make at a given game state in soccer. Computer vision was then added to support the import of images to analyze a game state.

Course Faculty

Anton Dahbura

Project Mentors, Sponsors, and Partners