Olympique de Marseille Recruitment System

This project brought advanced data science and machine learning to professional football recruitment for Olympique de Marseille. By ingesting and analyzing global match data from StatsBomb, I built a comprehensive system that transformed raw event data into actionable player insights.

The system provided OM's sporting director and recruitment team with data-driven transfer lists, moving beyond traditional scouting methods to quantify player value through probabilistic modeling.

Traditional football recruitment relies heavily on subjective scouting reports and basic statistics that fail to capture the true value of player actions. Clubs needed a way to:

• •Quantify the impact of every on-field action beyond goals and assists
• •Compare players across different leagues and positions objectively
• •Identify undervalued talent in markets beyond major leagues
• •Make data-driven transfer decisions with statistical confidence

VAEP (Valuing Actions by Estimating Probabilities) represents a paradigm shift in player evaluation. Instead of counting events, it measures impact:

• •Probabilistic scoring: Each action is evaluated by how it changes the probability of the team scoring or conceding in the near future
• •Context-aware: A pass in midfield is valued differently than the same pass in the attacking third
• •Comprehensive coverage: Every touch, movement, and defensive action contributes to a player's overall value
• •ML-powered: Machine learning models trained on thousands of matches learn what actions actually lead to goals

This approach revealed players who excel at "winning actions" - the subtle plays that increase scoring probability but don't show up in traditional statistics.

The system provided Olympique de Marseille's recruitment team with:

• •Custom transfer shortlists ranked by VAEP value and filtered by budget, age, and position needs
• •Quantitative player comparisons that revealed undervalued talent in less-scouted leagues
• •Data-driven insights to support or challenge traditional scouting assessments
• •Evidence-based negotiation positions for transfer discussions

This project demonstrated how machine learning can augment traditional football scouting, providing recruitment teams with objective, data-driven insights to complement human expertise.

I wrote a comprehensive three-part series on Medium exploring football analytics in depth, covering the methodology, implementation, and real-world challenges of this project: