VAEP Football Analytics: The Complete Guide to Valuing Every Action

Picture a midfielder who barely shows up on the scoresheet. No goals. Few assists. But watch closely and you will see someone who consistently makes the right decisions under pressure. Interceptions that snuff out counter-attacks. Passes that break defensive lines. Tackles in dangerous areas that prevent certain goals.

Traditional stats ignore this player. Bookmakers undervalue their team. Yet their contributions are worth serious money to bettors who know where to look.

That is where VAEP (Valuing Actions by Estimating Probabilities) comes in. Unlike goals, assists, or even Expected Goals, VAEP puts a number on every single on-ball action based on how it shifts the probability of scoring or conceding. For anyone serious about football analytics, this metric reveals what the market misses.

VAEP is an advanced football analytics framework created by the DTAI Sports Analytics Lab at KU Leuven and later commercialized by SciSports. The concept is simple but powerful: How much did this specific action change the game?

The system uses machine learning to evaluate every pass, tackle, dribble, clearance, and shot. Each action gets a value based on two probabilities:

• Scoring probability: The chance of scoring within the next 10 actions
• Conceding probability: The chance of conceding within the next 10 actions

What makes VAEP different from simpler football analytics metrics is its dual focus. It captures offensive contribution (boosting scoring chances) and defensive contribution (cutting conceding chances) in a single framework.

"VAEP is a probabilistic model that evaluates how a player's action affects the team's chances of scoring or conceding in the near future. It looks beyond basic statistics like goals and assists to capture the broader influence of actions."
— The Football Analyst

For bettors, this means access to a metric that finally gives credit to players who contribute in ways that never appear on a match report.

Traditional football statistics have a massive blind spot. Goals, assists, and even Expected Goals only capture a fraction of what happens on a pitch. A defender can dominate a match without getting anywhere near the scoresheet.

Take Expected Goals, the most popular advanced metric in football. xG only values shots. A defender who makes five crucial interceptions, a midfielder who plays defense-splitting passes that just miss their target, or a winger whose pressing forces errors all day long. None of these register in xG.

VAEP solves this by valuing all 16+ action types tracked in modern event data:

• Passes and crosses
• Dribbles and carries
• Tackles and interceptions
• Clearances and recoveries
• Fouls and ball losses
• Shots and assists

The result is a much fuller picture of what players actually do. Research from the DTAI lab shows VAEP has a novelty score of 0.21. Roughly 21% of its variance cannot be explained by 163 traditional performance metrics combined. That is genuinely new information for building a sports betting model.

The math behind VAEP involves sophisticated machine learning, but the underlying concept is straightforward enough to apply in practice.

Two gradient boosted binary classifiers trained on historical match data power the system. One predicts scoring probability within the next 10 actions. The other predicts conceding probability in the same window.

VAEP Formula:

V(action) = Change in Scoring Probability - Change in Conceding Probability

The model looks at the three most recent actions to understand context, analyzing 36+ variables including:

• Action characteristics (type, result, body part used)
• Location data (start/end coordinates, distance and angle to goal)
• Movement data (distance covered, time elapsed)
• Contextual features (current score, time remaining, home/away status)
• Advanced metrics like pitch control delta and defender pressure

Here is a concrete example. A midfielder plays a progressive pass from the center circle to the edge of the opponent's box:

• Scoring probability before: roughly 0.6% (possession in safe area)
• Scoring probability after: roughly 5% (possession in dangerous area)
• Conceding probability change: minimal
• VAEP value: roughly +0.04

"A progressive pass that breaks the opponent's midfield line and moves the ball closer to goal might add +0.05 to the scoring probability and -0.01 to the conceding probability. VAEP = 0.05 - (-0.01) = +0.06"
— The Football Analyst

Being able to quantify previously invisible contributions is exactly what makes VAEP valuable for finding market inefficiencies.

Several possession value frameworks exist in modern football analytics. Knowing the differences helps you pick the right tool.

Expected Threat (xT)

Created by Karun Singh, xT divides the pitch into zones with assigned values based on historical scoring rates. Players get credit for moving the ball to higher-value zones.

Advantages: Simple, interpretable, stable game-to-game
Limitations: Does not value defensive actions, ignores context like pressure and score differential

For a deeper dive, see our complete guide to Expected Threat (xT) in football analytics.

StatsBomb On-Ball Value (OBV)

Built on StatsBomb's xG model, OBV includes both goals for and goals against components. It deliberately excludes possession history features to avoid bias toward players on stronger teams.

Advantages: Reduces team strength bias
Limitations: Proprietary, not widely available outside StatsBomb clients

Stats Perform Possession Value (PV)

Uses a 10-second time window rather than counting actions. Originally punished ball losses heavily, but revised to reward even unsuccessful actions that end in dangerous areas.

Advantages: Time-based approach may capture game flow better
Limitations: Less academic validation than VAEP

Which Should Bettors Use?

The research comparison between xT and VAEP reveals something counterintuitive. xT achieves correlation of 0.89 for position-based values. VAEP achieves only 0.25. That lower correlation is actually a strength.

"The entire goal of innovation is to capture new information that current metrics miss. Therefore, a high correlation with existing metrics should be viewed with skepticism, not as a badge of honor."
— DTAI Sports Analytics Lab

VAEP's lower correlation reflects greater contextual complexity. For bettors, this means VAEP is more likely to surface value the market has not spotted yet.

For a comprehensive overview of all possession value frameworks, see our guide to possession value models in football analytics.

The practical value of VAEP for sports betting analytics comes down to one thing: revealing information invisible to standard analysis.

Identifying Undervalued Players with VAEP

The top VAEP performers in the 2024/25 season (excluding shots, minimum 900 minutes) included:

1. Ousmane Dembele (PSG) - 0.41 VAEP/90
2. Trent Alexander-Arnold (Liverpool) - 0.38 VAEP/90
3. Luka Modric (Real Madrid) - 0.36 VAEP/90
4. Hakan Calhanoglu (Inter Milan) - 0.35 VAEP/90
5. Michael Olise (Bayern Munich) - 0.35 VAEP/90

Notice the mix. Creative midfielders. Attacking full-backs. Players whose value extends far beyond goals and assists. Someone like Modric, with declining traditional output but elite VAEP, represents exactly the type of market inefficiency analytical bettors can exploit.

Predictive Validation of VAEP Football Analytics

VAEP has proven it can predict future performance. The framework correctly identified future stars in Eredivisie U21 football, including Frenkie de Jong (whose market value jumped from EUR 7m to EUR 90m), Mason Mount, Donny van de Beek, and Steven Bergwijn before they became household names.

A Bundesliga study in 2025 found that Expected Possession Value models achieved 58.3% prediction accuracy compared to 55.6% for xG alone in pre-match scenarios. That 3 percentage point improvement matters in markets where margins are thin.

The Pre-Assist That Traditional Stats Ignore

January 2019. Aymeric Laporte plays a through ball to set up a goal for Manchester City against Wolverhampton. Traditional statistics credit the assist (Leroy Sane) and the goal (Gabriel Jesus). VAEP assigns a high positive value to Laporte's "pre-assist" too, the pass that made the assist possible.

For bettors analyzing player props or team performance, recognizing these hidden contributions can reveal value in markets focused on specific players.

The Risky Back Pass

A defender plays a short back pass near their own goal. VAEP assigns this action a negative value because:

• Scoring probability drops (ball moved away from goal)
• Conceding probability may rise (possession in a dangerous area)

This risk-reward analysis helps bettors spot teams that take unnecessary risks at the back, valuable for goals markets and clean sheet betting.

The Crucial Tackle

A defender makes a sliding tackle on the edge of their own box. VAEP assigns a highly positive value because the action slashes the opponent's scoring probability while potentially boosting the team's own scoring chances through regained possession.

This proper valuation of defensive actions lets bettors identify defenders whose contributions go far beyond what clean sheets and tackles won statistics capture.

The Failed But Valuable Action

A cross from Andrew Robertson does not reach its target. But the clearance falls to a teammate in a better position. Modern possession value frameworks (like Stats Perform's updated PV) assign positive value (+0.01) to this action because the ball ended up in a better spot within 10 seconds.

For bettors, this shows how even unsuccessful actions can contribute to team performance. Important when analyzing teams that create chances without converting them.

1. Identify Undervalued Teams with VAEP Data

Look for teams with high aggregate VAEP but low goal conversion. These sides create dangerous situations more often than their goal totals suggest. They may be due for positive regression. This approach works particularly well in long-term markets like points totals and relegation battles.

2. Match Prediction Enhancement Using Football Analytics

Feed team-level VAEP aggregations into match outcome models. The Bundesliga study showed possession value metrics can improve prediction accuracy by 3+ percentage points compared to xG alone. In a market where margins are razor-thin, that improvement matters.

3. Live Betting Opportunities with Real-Time VAEP

Monitor possession value momentum shifts during matches. Teams showing consistently high VAEP without converting may be about to score. Value appears in next goal markets or in-play handicaps. The Belgium vs Japan 2018 World Cup match illustrated this perfectly. Japan showed positive momentum before Belgium's comeback began. For more on in-play wagering, see our complete guide to live betting.

4. Player Prop Markets and Advanced Statistics

High-VAEP players get involved in dangerous situations more often. This translates to value in markets like:

• Player shots on target
• Player touches in opposition box
• Player passes into final third

For a deeper dive into profiting from these markets, see our guide to player shot markets.

5. Transfer Impact Analysis Using VAEP

Track VAEP trends for players involved in transfers. Teams acquiring high-VAEP players (especially undervalued ones) may outperform expectations. Teams losing their VAEP leaders may decline even if the departing player was not a prolific scorer.

6. Risk-Reward Profiling in Football Betting

Identify teams with high-risk, high-reward playing styles through their VAEP profiles. Teams with many players attempting high-value but risky actions may have higher variance in performance. Relevant for over/under markets and handicap betting.

7. Tactical Matchup Analysis with Advanced Metrics

Look at where teams generate VAEP value. Teams that create value primarily through wide areas may struggle against packed defenses. Teams with high defensive VAEP may be stronger underdogs than traditional stats suggest.

"VAEP helps us understand how each action influences a team's chances of scoring or conceding, going beyond the obvious."
— SportBot AI

No metric is perfect. VAEP has important limitations bettors need to understand.

Off-Ball Actions Not Captured by VAEP

VAEP only values on-ball actions. Off-ball movement, positioning, pressing without contact, and defensive shape remain invisible. A striker whose run pulls defenders away gets no VAEP credit, even though they create space for teammates.

Data Quality Dependency in Football Analytics

VAEP relies on accurate, detailed event data. Different data providers may produce different VAEP values for the same match. Not all leagues have high-quality event data available.

Black Box Elements in VAEP Models

Despite using interpretable features, explaining why a specific action received a particular value can be tricky. As one analyst noted: "VAEP can behave like a black box. Precisely explaining why a particular action received a certain value is often not straightforward."

Game-to-Game Variability in VAEP Ratings

VAEP shows moderate robustness but is not incredibly stable from match to match. Shots receive high values but occur infrequently with variable outcomes. Some analysts exclude shots from aggregate ratings to increase stability.

Team Strength Bias in VAEP Analysis

Players on stronger teams tend to accumulate more positive VAEP simply because their teams are in advantageous game states more often. Some frameworks address this by excluding possession history features.

When comparing VAEP vs xG, the key distinction lies in scope. Expected Goals measures shot quality, the probability that a shot becomes a goal. VAEP measures action value, how any action (not just shots) affects the probability of scoring or conceding.

Feature	Expected Goals (xG)	VAEP
Actions valued	Shots only	All on-ball actions
Defensive contribution	Limited	Full defensive value
Build-up play	Not measured	Every action valued
Scope	Final third	Entire pitch
Risk assessment	No	Yes (conceding probability)

For comprehensive sports betting analytics, using both metrics together works best. xG tells you about finishing quality. VAEP tells you about overall contribution.

VAEP represents a real step forward in football analytics. It gives bettors a more complete picture of player and team performance than traditional metrics allow. By valuing all on-ball actions based on their impact on scoring and conceding probabilities, VAEP reveals contributions that goals, assists, and even xG miss entirely.

For bettors seeking an edge, VAEP provides tools to identify undervalued players, enhance match prediction models, and spot momentum shifts during live betting. The framework has real validation behind it, from identifying future stars to improving prediction accuracy.

However, VAEP works best as part of a broader analytical toolkit. Its limitations around off-ball actions, data quality requirements, and interpretation complexity mean it should complement rather than replace other analysis methods. Used wisely, VAEP can help bettors find value in places the market has not looked yet.

The most successful football betting strategies combine multiple data sources with contextual understanding. VAEP offers a powerful lens for viewing player contribution, one that can reveal market inefficiencies when applied thoughtfully to betting decisions.

• Possession Value Models: Football Analytics Guide for 2026
• Expected Threat (xT) in Football: Complete Analytics Guide
• How to Build a Sports Betting Model: Complete Guide
• Live Betting Explained: Complete In-Play Betting Guide 2026
• Player Shot Markets: How to Profit from Shots on Target