Match Analysis: Performance Indicators That Actually Predict Success

Introduction

Football is awash with data. Every professional match generates millions of tracking data points, hundreds of event codes, and dozens of derived metrics. But data abundance creates a new problem: identifying which performance indicators actually predict winning, rather than merely correlating with successful teams in circular ways. The last 20 years of match analysis science have progressively refined which numbers matter and the answer challenges some long-held coaching assumptions.

The Science

Performance indicators (PIs) in football analysis are measured variables that discriminate between winning and losing teams, or between elite and sub-elite performance. The challenge: most obvious metrics (possession percentage, total shots, total distance) have surprisingly weak relationships with winning across large datasets.

What the data consistently shows:

Shots on target and specifically expected goals (xG) are the strongest match outcome predictors. xG assigns a probability to each shot based on location, shot type, assist type, and game-state context, derived from historical conversion rates across thousands of shots. Team xG outperforms simple shot counts as a predictor of outcome because it weights shot quality, not just frequency. Over a season, xG performance strongly predicts final league position.

High-intensity pressing metrics: PPDA (Passes Allowed Per Defensive Action) measures the average number of passes an opponent completes before a defensive action (tackle, interception, pressure) is made. Lower PPDA = more intense pressing. Research confirms that PPDA is one of the strongest team-level predictors of success in modern European football, reflecting the tactical-physical integration that defines pressing systems.

High-speed running and sprint distance, particularly in the final 30 minutes of matches: shows significant correlations with winning in European league data, supporting the fatigue-management principles discussed in Article 7.

What does not reliably predict winning:

• Possession percentage: Teams can win with 35% possession (Leicester 2015–16) and lose with 70% possession routinely. Possession without shots on target is decorative.
• Total distance covered: Broadly similar across winning and losing teams when controlled for match context.
• Total passes: Volume without quality or territory is noise.
• Tackles won: Weakly predicts outcomes; the causal direction is ambiguous (teams losing more often need to tackle more).

What Research Says

Lago-Peñas et al. (2010) analysed 380 matches from the Spanish La Liga in Journal of Sports Science and Medicine, finding that shots on target, shots in the penalty area, and high-speed running in the attacking half were the variables that most strongly discriminated between winning and losing teams. Possession percentage did not significantly discriminate.

Rein and Memmert (2016) published a comprehensive review of big-data approaches to football analytics in European Journal of Sport Science, confirming that spatial analysis (where shots come from, where defensive actions occur) adds significant predictive value beyond simple event counts. The development of pitch control models and possession value frameworks represents the current frontier.

Gegenpressing-focused analysis from Fernández et al. (2018), drawing on tracking data from La Liga, demonstrated that defensive pressure intensity (PPDA) and ball-recovery location within 40m of the opponent’s goal were significant predictors of subsequent scoring opportunities. The pressing data validated Klopp’s claim that “the best playmaker in the world is a good press”, in statistical terms.

Did You Know? When Liverpool hired data analytics firm StatDNA (later acquired by Arsenal) and subsequently developed their own data science department, they identified that xG-differential (shots on target quality adjusted for chance quality) predicted league position more accurately over 5+ years than any traditional metric. The signing of Mohamed Salah was partly driven by his exceptional xG-adjusted scoring record in Serie A, his underlying numbers suggested systematic underperformance of his actual goal output relative to his chance quality. Liverpool’s model predicted his Premier League conversion before most scouts agreed.

Applied to Football

Applying performance indicator science to club practice:

1. Track xG, not just goals. Expected goals reveals whether a team’s results match their underlying performance. A team winning with negative xG differential is outperforming their expected results and will regress. A team losing with positive xG is underperforming and likely to improve. xG is a more stable indicator of performance level than goals.
2. Use PPDA to quantify pressing intensity. Clubs can calculate approximate PPDA from available tracking data. Setting PPDA targets for each team shape (a 4-2-3-1 pressing team vs. a 4-5-1 defensive team will have very different targets) gives pressing intent a measurable, trackable form.
3. Analyse shot locations, not just shot counts. Shots inside the 6-yard box have 50%+ conversion; shots from outside the box average 3–5%. A team taking 18 shots from outside the box is performing worse than a team taking 8 shots in the penalty area despite the raw shot count advantage.
4. Be critical of possession statistics. High possession without territorial control or shot creation is a vanity metric. Track what happens at the end of possession sequences, do they end in shots, dangerous areas entered, or wasteful lateral passes?
5. Contextualise physical data with match context. A team chasing a 2-0 lead in the 80th minute will have higher HSR, but this is reactive, not indicative of superior fitness. Match state context is essential for meaningful physical data interpretation.

In Short

• xG (expected goals) is the strongest match-level predictor of future outcomes; it outperforms raw shot counts
• PPDA (Passes Per Defensive Action) quantifies pressing intensity and correlates with success in modern tactical football
• Possession percentage does not reliably predict winning, shot quality and territory matter more than possession volume
• Shots inside the six-yard box have 10–15× higher conversion rates than shots from outside the box, location dominates volume
• Match state context (score, time elapsed, opposition quality) is required for meaningful physical data interpretation

References

• Lago-Peñas, C., Lago-Ballesteros, J., Dellal, A., & Gómez, M. (2010). Game-related statistics that discriminated winning, drawing and losing teams from the Spanish soccer league. Journal of Sports Science & Medicine, 9(2), 288–293.
• Rein, R., & Memmert, D. (2016). Big data and tactical analysis in elite soccer: future challenges and opportunities for sports science. SpringerPlus, 5(1), 1410.
• Fernández, J., Bornn, L., & Cervone, D. (2018). Decomposing the immeasurable sport: a deep learning expected possession value framework for soccer. MIT Sloan Sports Analytics Conference.

Next in Series: The Physiology of Pressing: Why High Pressure Demands Extraordinary Fitness

Written by Hüseyin Akbulut, MSc Coach

Hüseyin Akbulut is the founder of Sporeus and author of THRESHOLD (EŞİK), a 540-page Turkish-language book on endurance science.

• 540pp THRESHOLD Book
• MSc Sport Sciences
• Marmara University

Full profile →