Rethinking How We Measure Soccer Performance

2026-07-08 · Source: Knowledge at Wharton · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, extended

Summary

Jonathan Pipping, a PhD student at Wharton, introduces XG+, a novel soccer analytics metric designed to overcome the limitations of traditional expected goals (XG). XG+ accounts for both the probability of a shot being taken and its likelihood of becoming a goal, addressing dangerous opportunities that XG misses. This new approach leverages continuous player tracking data, estimated 30 times per second, to provide a more complete picture of an attack. XG+ also resolves issues like accumulating over one expected goal from rebounded shots. Empirical findings show XG+ metrics, particularly "expected shots," are more stable year-to-year for player performance evaluation than XG overperformance, with a correlation of 0.6 or a 0.7 compared to XG's 0.1. This makes XG+ a more predictive tool for assessing player skill and transfer value.

Key takeaway

For sports analysts and team managers evaluating player talent, XG+ offers a more stable and predictive metric than traditional XG. You should prioritize players who consistently overperform their "expected shots" as this skill is more likely to carry over to future seasons, preventing overpayment for past, less reliable finishing streaks. Implement XG+ to identify market inefficiencies and acquire players whose underlying skill in creating opportunities is undervalued.

Key insights

XG+ enhances soccer analytics by modeling both shot probability and goal likelihood, revealing missed dangerous opportunities.

Principles

• Player skill in creating shooting opportunities is more predictive than finishing ability.
• Overperformance on XG is noisy; XG+ metrics are stickier and more consistent.

Method

XG+ estimates the probability of a shot occurring from a given configuration (players, ball, goal openness) and combines it with the conditional probability of scoring once a shot is taken.

In practice

• Evaluate player transfer value using XG+ to identify consistent, skill-based performance.
• Assess team offensive threat by quantifying dangerous attacks that do not result in shots.

Topics

• Soccer Analytics
• Expected Goals
• XG+
• Player Tracking Data
• Sports Statistics
• Player Evaluation
• Transfer Value

Best for: AI Scientist, AI Student, Data Scientist, Research Scientist

Related on AIssential

• Blame is easier than praise: Measuring off-ball defensive performance in football — A novel framework quantifies off-ball defensive performance in football by attributing threat changes to individual players' positional behavior, even without ground truth.
• How $5.8 Billion in Prediction Markets Are Pricing the World Cup Final — Prediction markets provide more efficient, real-time event pricing than traditional sportsbooks by eliminating house margins and reacting faster to information.
• World Cup Teams Are in a Race for AI Dominance — The World Cup's unprecedented data scale fuels an AI competition among teams for performance advantage.
• Can AI really spot the next football superstar – or is it changing the game in troubling ways? — AI in football scouting risks perpetuating biases by relying on historical data and may overlook unpredictable talent development.
• Why Barcelona Chose Anthony Gordon Over Marcus Rashford: A Data-Driven Look — Barcelona's transfer decision prioritized long-term financial sustainability, defensive system fit, and asset value over immediate goal output or reputation.
• How Qatar Became FIFA’s Technology Test Lab — Qatar served as a real-world testbed for FIFA's advanced football technologies, accelerating their global deployment.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Knowledge at Wharton.