Football Player Potential Prediction Model

Scoutium's scouts assess players from detailed attribute data, but judging genuine potential by eye is subjective and hard to keep consistent across a large pool of prospects. Two scouts can weigh the same attributes differently, and there was no objective, repeatable read on how much talent a given player profile actually implied.

Scoutium wanted a data-driven second opinion: a model that could look at a player's attributes and estimate their talent level consistently, to support - not replace - the scout's judgement.

I built a machine-learning model that predicts a player's talent level directly from their scouted attribute data. I engineered features from the raw attributes, trained and tuned a classification model, and validated it so its accuracy would hold on players it had not seen during training.

The finished model gives scouts an objective, repeatable read on potential with roughly 85% accuracy, turning a large table of attributes into a single, comparable talent signal that sits alongside human scouting judgement rather than overriding it.