Skip to Main Content
Team Performance

Quantifying Player Contribution to Counter Attacks


In a guest blog for Stats Perform, Laurynas Raudonius presents his findings from a research project which applied tracking and event data with the objective of quantifying each individual player’s contribution to a team’s counter attacks.


By: Laurynas Raudonius

Laurynas Raudonius was the inaugural winner of the Dr. Garry Gelade award, which recognises the outstanding proposal submission from an undergraduate at Stats Perform’s Pro Forum.

After winning the award, Laurynas exhibited a poster presentation at the virtual 2021 event, which illustrated the findings of a project which applied spatio-temporal data to quantify how individual players contribute to a team’s successful counter attack, using Voronoi cells and other methods.

Click here to view his poster.

Eliminating Subjectivity When Assessing Counter Attacks

Why is it important to be able to objectively quantify each player’s contribution to a counter attack?

A good starting point would be to challenge yourself, using just your naked eye, to identify and award player scores to the most important contributions from a single isolated counter attacking scenario.

When you watch the following video, consider what player contributions were the most impactful, and why.

Having watched this clip, I am sure you will now appreciate the challenge of objectively assigning numerical values to actions performed by footballers. This is, without doubt, one of the great challenges of football analytics.

The benefits of creating a model to achieve that are obvious – having an unbiased comparison of any two players can help enormously in scouting, team selection and many other areas of football.

My model combines four metrics that are calculated using tracking and event data and applies them to place a value on player contributions to counter attacks. Whilst there are many existing football-related studies which focus on attributing value to passes, this model takes things a step further and values all actions by combining them into contributions. Somewhat similar models have been developed, but they have been trained as Machine Learning models and do not include many separate indicators.

Furthermore, this project focuses on counter attacks, a subset of attacking scenarios that have received very little attention to date from data scientists and researchers.

Defining a Contribution

The model used in this project does not differentiate between different types of attacking actions made by players, but instead combines them all into contributions. A contribution is essentially a pair of in-game possession states (the position of all the players and the ball on the pitch at a moment in time), starting firstly with the moment when a player first touches the ball and secondly the moment a team mate subsequently touches the ball (and, in turn, starts their own contribution). The difference between these two in-game possession states answer the question ‘how did a player impact the situation whilst on the ball?’

Deriving the Exact Score of a Contribution

The difference between the two possession states won’t tell us a great deal unless we have a way of objectively assigning numerical values to them. For this, four separate metrics, derived from tracking and event data, have been designed and implemented:

1. Distance

This is the most intuitive metric of the four. It is obvious that if a player carries the ball from, for example, their own third to the opposition’s penalty box, they have contributed a lot to making the attack more dangerous. Therefore their contribution should have a high valuation. To measure this exactly, the model calculates the Euclidean distance between the ball and the goal when a player first obtains the ball and the same distance when they finish their action (i.e. attempt a pass). The difference between the two tells us how much closer the ball is to the goal as a result of the player’s involvement and therefore is the value of the indicator. Negative values can also be assigned as it is possible that a player can take the ball further away from goal.

2. Danger

Building on the distance metric, we can account for the exact location on the pitch the ball is following the player’s actions, not just how far it is from the goal.

Assessing the danger, of different areas of the pitch, is a topic that has been investigated in-depth by Daniel Link et al. in their research on danger in football attacks. Their research proposes to divide the attacking third into 2×2 meter squares and then assign a danger score between 0 and 1 to each square (see Figure 1 below).

To calculate the exact values they follow five key rules:

a) As the distance from the goal decreases and centrality increases, the danger rises.

b) Moving into the penalty area brings about a sudden increase in the danger because of the risk of the defensive team conceding a penalty kick.

c) There is a homogeneous area in front of the goal in which the danger does not increase further.

d) An acute angle to the goal reduces the danger.

e) Areas to the side of the penalty area are dangerous because of the possibility of a cross with little risk of offside.

3. Outplayed Players

Whilst the distance and danger metrics are informative, they are fundamentally based on the ball position only.

What if a player had a clear path to goal from the halfway line? Based on the previous two indicators alone, the contribution of that player would be valued greatly, even though they were under little pressure from opposition players. Therefore it is important to consider, when valuing player actions, how they impact the opposition defence.

This is exactly what the third indicator measures. To be more specific, it calculates how many opposition players are behind the ball during the player’s actions. Since goalkeepers stay behind the ball most of the time, they are not accounted for in this metric.

4. Space Control

A study has explored the relationship between a team’s success and how much space they controlled in the opponent’s half, 30 metres from goal. It found there was a somewhat direct correlation: many successful teams, such as Barcelona and Borussia Dortmund in the early 2010s, controlled large spaces in their opponents’ half, so it only makes sense to award a higher valuation to players whose contributions increase the controlled space.

For this project, the controlled space is measured using Voronoi cells. A player’s Voronoi cell is the set of points on the pitch they are closer to than any other player.

Calculating space across the whole pitch would not be very informative, because a player who increases the space in their own half does not necessarily contribute to a counter attack. Therefore a threshold (area) on the pitch has been selected, determining when space control is accounted for.

Experimentation with different areas in front of the goal indicated that if controlled space is measured within the final quarter of the pitch only, then variance in space control is greater across all the players, allowing us to better differentiate the significance of a player’s contribution. Therefore this was set as a threshold in the model.

Combining the Metrics

In order to produce a single score per player contribution, the four indicators have been combined. The first step is to normalise them all into the same range: here [-1;1] was selected. Then the scores for a contribution are added up and multiplied by 2.5 to bring the obtainable scores between -10 and 10. This then results in our final score assigned to a contribution.


Having established the methodology, we now need to demonstrate how the model can be applied to an arbitrary passage of play in a match.

If a passage is relatively uncomplicated (such as Gareth Bale’s winning goal from a counter attack in the 2014 Copa del Rey) the challenge of distributing scores for contributions is rather straightforward, however it gets exponentially harder with every new player who joins in with the play.

This model is capable of solving that problem.

Let’s return to the video example we showed at the start of this piece. In this new version of the clip, we pause the footage at the end of each contribution and display how the model valued it.

The bar chart below summarises the values that were assigned to contributions during this counter attack.

When you initially considered what player contributions were the most impactful in this specific counter attack, it would be interesting to know if your values were similar to those derived by this model.

Now that I have explained my methodology to you, for a bit of fun let’s do this same exercise again but this time, try to take into account the four indicators when you allocate the scores to each contribution.

Once you have watched the clip, click on the video below to see the contribution values attributed by the model and compare them to your own.

Again, the following bar chart summarises the values of each counter attack contribution.

In this example, the contribution of player six is worth highlighting. His contribution, according to the model, is the largest, with a combined score of almost five. However when I showed this video to delegates at the Pro Forum, people tended to say that they thought player eight or player nine made the highest contribution. This reinforces the subjectivity of trying to quantify a player’s contribution to counter attacking phases using just the naked eye.

It also further highlights the potential of this type of research to try and reduce bias when assessing recurring match scenarios, which can potentially enhance processes in match analysis and inform recruitment profiling.


Tavares, Ricardo. (2019). Using Voronoi Diagrams in Football.

Link, D., Lang S., and Seidenschwarz P. 2016. Real Time Quantification of Dangerousity in Football Using Spatiotemporal Tracking Data.

Rein, R., Raabe D., and Memmert D. (2017). “Which Pass Is Better?” Novel Approaches to Assess Passing Effectiveness in Elite Soccer. Human Movement Science 55.

Perl, J. & Memmert, D. (2017). A Pilot Study on Offensive Success in Soccer Based on Space and Ball Control – Key Performance Indicators and Key to Understand Game Dynamics.

Originally from Lithuania, Laurynas Raudonius graduated in Computer Science at The University of Manchester earlier this year. Now based in Switzerland, he has previously worked as a match analyst for Lithuanian top-flight side FK Kauno Žalgiris.