Defensive Responsibility: A New Way To Measure Defensive Output
4 Jun 2026
7 min Read
Defending is one of the most complex and least understood aspects of football analytics. We can measure what a team concedes - expected goals against, possession value conceded - but attributing that to individual players has always been far more difficult.
For years, trying to quantify individual defensive contributions from event data has been an uphill battle. The earliest approaches simply counted raw defensive actions, such as tackles, interceptions, and blocks. But it quickly became clear that raw counts don’t measure defensive quality; they only measure defensive volume. A centre back in a low-block team racked up the blocks and clearances, while world-class defenders were entirely anonymous on a spreadsheet.
Possession-adjusted metrics were the next step forward, normalising defensive actions against the team’s time in possession. We could account for the fact that a player on a low-possession team faces more defensive situations than one on a team that dominates the ball. Now we’re cooking… or so we thought.
It was a meaningful improvement, but it introduced a different problem in treating every moment of opposition possession as an equal defensive opportunity, regardless of where the ball is, what kind of action is being played, or which player on the defending team would actually be expected to respond to it. A lateral pass between two centre backs in the opposition half counts the same as a cutback across the six-yard box.
To truly understand defensive impact, we have to stop treating defending as a collection of isolated events, and start treating it as a dynamic system. We need a model that understands not just what happened, but who was actually supposed to deal with it.
Enter Defensive Responsibility (DefR).
Defensive Responsibility (DefR) is an event-based framework that distributes defensive accountability across a team. It looks at every carry, pass, and shot the opposition makes and predicts - based on the location of the action, the momentum of the attack, and the defensive shape of the team - which player in which role would typically be expected to respond to it.
The result is a Responsibility Matrix. For every player, across every match, we know how many defensive actions were expected of them and how many they actually made. The difference - defensive actions above or below expectation - is the signal.
So what does DefR actually tell us that we didn't know before? The clearest way to answer that is to put it side by side with possession-adjusted defensive actions and see where the two metrics disagree.
The two metrics do correlate, but the divergences are where it gets interesting.
Take Kylian Mbappé. Real Madrid averaged 59% of the possession in 2025/26, lifting the possession-adjusted numbers of their entire squad. PAdj sees Mbappé as a modestly infrequent defensive contributor, but DefR is more sceptical, identifying Mbappé as the clear outlier when it comes to defensive output - accounting for the actual attacking actions played through his zone and his defensive output against them. His defensive activity map also shows this.
On the other hand, by possession-adjusted metrics Adrián Liso of Getafe looks like a middle-of-the-pack defensive contributor. But Getafe average just 41% possession, meaning PAdj has been quietly depressing the defensive numbers of Liso and his teammates, while underselling what Liso is doing relative to what was asked of him. DefR corrects for that.
This is the core upgrade DefR offers over PAdj: it stops penalising players for their team's lack of the ball, and stops rewarding players for their team's dominance of it.
DefR doesn't just transform how we evaluate individual players, it changes how we read teams too.
At the team level, the metric tells us which teams are defending more aggressively than the situations they faced demanded, and which are falling short. The difference from PPDA is that DefR accounts for the specific nature of the attacking actions a team faced, rather than treating every opposition pass as an equal defensive opportunity.
Brighton top the chart, with Newcastle close behind - two clubs whose managers, Fabian Hürzeler and Eddie Howe, have built sides to be aggressive and disciplined off the ball. At the other end, West Ham and Sunderland’s more passive defensive approach, preferring to drop into a shape and contain space rather than actively engaging the opponent, is reflected clearly in the data.
Applying Defensive Responsibility (DefR)
Evaluating centre-backs is notoriously difficult because their data is highly dependent on their team's defensive approach and the opponent's attacking approach. By plotting a defender's Expected Defensive Actions against their Actions Above/Below Expectation, we can filter out team noise and divide centre-backs into four distinct tactical profiles.
The Front-Foot Aggressor
Top-Right Quadrant: High Expected Actions / Actions Above Expectation
Bournemouth’s Marcos Senesi sits firmly in this quadrant. His system demands a high defensive workload, and Senesi consistently meets them, stepping out of the defensive line to aggressively shut down attacks before they develop.
The System Absorber
Bottom-Right Quadrant: High Expected Actions / Actions Below Expectation
Castello Lukeba at RB Leipzig faces a high volume of expected responsibility based on his position and the attacking approach of RB Leipzig’s opponents, but he performs fewer defensive actions than expected. This implies that he relies more on his recovery pace and positional containment within Leipzig's structure than meeting and engaging with the play, but this is something that thorough video scouting should verify.
Aston Villa’s Ezri Konsa occupies this space - as does his centre back partner Pau Torres. Both play in a highly disciplined, structured defensive unit where their primary role is to hold the offside line and protect the penalty box rather than hunting for the ball.
The most fascinating quadrant is, arguably, in the top-left.
The common narrative around Virgil van Dijk is that he is a more passive defender who simply shepherds attackers away from danger without needing to make a tackle. The data tells a much more nuanced story. Van Dijk's low expected output confirms that Liverpool are effective at preventing their opposition from playing through his defensive zone (or, rather, teams deliberately steer their attacking plays away from his zone knowing that he is best avoided). However, DefR reveals that when opponents do dare to play through his territory, van Dijk is highly efficient, consistently performing more defensive actions than the model expects. Perhaps he isn't passive after all, more just ruthlessly selective.
For a recruitment analyst, these quadrants can help to identify players who best fit the defensive demands of your team and the specific profile of centre back that would work most effectively.
Central Midfielders, though having different demands placed upon them, can be evaluated similarly.
Comparing the output of Elliott Anderson and Pedri demonstrates the utility of DefR.
In raw numbers, Anderson’s role in a low-possession Forest side sees him engage in slightly more tackles, interceptions, and pressures per 90 minutes than Pedri. When you apply the traditional possession-adjustment, the data swings wildly the other way: Pedri skyrockets to 35 PAdj Pressures per 90 - an alarming 70% higher than Anderson’s 22 PAdj Pressures per 90.
But DefR reveals that both metrics miss the true story. By measuring their output against the exact spatial and tactical opportunities they faced, DefR reveals that Anderson and Pedri are actually performing at an identical, elite level - registering roughly 8.2 defensive actions above expectation. Anderson is an ultra-active destroyer tasked with a high defensive workload who consistently goes above and beyond for his team; Pedri plays in an intensely aggressive Hansi Flick system where he actively hunts down the ball the second possession is lost. Traditional metrics forced a gulf between them; DefR recognises that their individual defensive hunger is exactly the same.
Our new DefR model isn’t a tool reserved exclusively for the elite. It is fully integrated across Hudl Statsbomb's global database, allowing analysts to scale their recruitment across different continents, tiers, and both the men's and women's games.
With DefR, recruitment analysts and clubs can finally cut through the noise of possession dominant and low-block biases to identify genuine out-of-possession workload. By moving away from flat normalisations and treating the defending team as a dynamic network of positional roles, DefR allows us to evaluate players based on what their specific tactical system and the phase of the opposition's attack actually demanded of them.
