Understanding Team Decision-Making Through Net Expected Threat (xT)

Modern football analysis has largely moved beyond raw possession and shot counts. What truly separates teams today is how their decisions with the ball shape the probability of future danger. This is where Expected Threat (xT) becomes one of the most powerful lenses available.

In this article, we break down a full match (or sample of matches) using net xT added, spatial aggregation, and action-level context to answer a simple but crucial question:

Where does this team create value, and where does it quietly lose it?

Rather than focusing on isolated highlights, the goal is to understand the structural tendencies behind a team’s attacking behavior.

Methodology: From Events to Spatial Value

The analysis is built on event-level data, where each on-ball action has:

• A start location

• An end location

• An associated xT value before and after the action

  
  

For each action:

xT_added = xT_end − xT_start

This allows us to evaluate:

• Whether an action increased or decreased the team’s attacking potential

• How value accumulates spatially over the pitch

  
  

To move from individual actions to team structure, events are aggregated into 24×16 pitch bins, covering the full field from:

• x: −52.5 to +52.5

• y: −34 to +34

  
  

For every zone, we compute:

• COUNT: how often actions occur

• MEAN xT added: average quality of actions

• SUM xT added: total accumulated impact

  
  

This separation is critical: volume, efficiency, and impact are not the same thing.

Where Actions Start: Territorial Habits

The starting-location count heatmap provides the baseline: where the team chooses to operate.

The distribution shows a team that:

• Uses a broad territorial footprint

• Engages frequently in middle and wide zones

• Does not rely exclusively on deep buildup or direct long balls

  
  

This suggests structural involvement across multiple lines rather than a single buildup corridor.

However, frequency alone is deceptive.

Start Locations – Average Quality

When we switch from volume to mean xT added, the picture changes.

Some high-frequency zones:

• Produce neutral or even negative average value

• Function more as circulation areas than progression hubs

  
  

Conversely, several less-used zones show consistently positive xT gains, indicating:

• Better decision-making

• More forward-oriented actions

• Higher contextual value

This immediately raises an important point:Not all possession zones are equally productive, even if they are frequently used.

Start Locations – Total Impact

The SUM xT map reconciles volume and efficiency.

Here we see:

• A small number of zones accounting for a disproportionate share of total attacking value

• Areas that are not spectacular individually, but become decisive through repetition

This is where tactical identity emerges: value is not only about brilliance, but about repeatable advantage.

Where Actions End: The Resulting Geography of Threat

Starting positions tell us about intention. Ending positions tell us about outcome.

End Locations – Frequency

Compared to starting locations, ending locations are:

• Slightly more advanced

• More concentrated near the attacking third

  
  

This confirms that the team is not only circulating possession, but progressing it.

Yet again, frequency alone does not guarantee effectiveness.

End Locations – Mean Value

Certain ending zones consistently generate high average xT gains.

These zones typically correspond to:

• Central pockets between lines

• Advanced half-spaces

• Wide-to-inside progression corridors

  
  

Importantly, not all advanced zones perform equally well. Some areas near the box still show neutral or negative averages, often linked to forced actions or low-quality deliveries.

End Locations – Accumulated Threat

This map reveals where the team’s attacks actually materialize into danger.

The highest-impact zones combine:

• Frequent arrivals

• Positive average outcomes

  
  

From a scouting or coaching perspective, these are priority zones:

• To defend against

• Or to deliberately reinforce in attacking patterns

  
  

Positive Value Only: Where the Team Gets It Right

So far, we have looked at everything together. Now we isolate only actions that add xT.

Positive xT – Start Locations (COUNT)

This map answers a powerful question:

Where do successful attacking actions actually begin?

The distribution is notably tighter than the general start map:

• Fewer zones

• Clear spatial preferences

This suggests that the team’s effective possession is far more structured than its overall possession.

Positive xT – Mean Value

Some zones show exceptionally high average gains:

• Even if they are used sparingly

• Often linked to decisive passes, carries, or switches

These are high-leverage areas:

• Small volume

• High reward

  
  

Positive xT – Total Contribution

Here we see the backbone of the attack:

• Zones that consistently fuel threat accumulation

• Areas where repetition and quality align

  
  

This is where tactical principles turn into measurable output.

Negative Value: Where Attacks Break Down

This is the most underused — and often most revealing — part of xT analysis.

Negative xT – Start Locations (COUNT)

These are the zones where actions frequently reduce attacking potential.

Two different patterns emerge:

1. Zones with high frequency and low value

2. Zones with low frequency but very costly mistakes

Both matter, but for different reasons.

Negative xT – Mean Loss

Some zones show strongly negative averages:

• Individual decisions here are particularly harmful

• Often associated with risky passes, poor body orientation, or pressure traps

These are decision-risk zones.

Negative xT – Accumulated Loss

This is the structural warning map.

Large negative sums indicate:

• Repeated suboptimal decisions

• Systemic issues rather than isolated errors

From a coaching standpoint, these zones demand intervention:

• Either through positional adjustments

• Or by modifying decision rules

  
  

Action-Level Context: High-Impact Moments

Finally, the scatter plot brings the analysis back to the human level.

Each circle represents an action:

• Size proportional to absolute xT impact

• Positive and negative decisions coexist spatially

  
  

This visualization reveals:

• Clusters of repeated high-impact behavior

• Isolated moments that swing attacking potential

• Areas where players face consistently difficult decisions

  
  

It is the bridge between model output and match reality.

What This Analysis Tells Us Beyond Goals

Several key conclusions emerge:

1. Possession geography and value geography are not the same

2. The team’s most frequent zones are not always its most productive

3. Positive value creation is spatially concentrated

4. Negative value accumulation exposes structural weaknesses

5. Decision-making quality varies sharply by zone

   
   

Most importantly, xT allows us to evaluate process, not just outcome.

Conclusion

Goals remain rare events.But decisions are constant.

By mapping where actions start, where they end, and how much value they add or destroy, we gain a far deeper understanding of a team’s attacking identity.

This approach does not replace video or tactical analysis — it guides it, pointing us to the zones, patterns, and decisions that truly matter.

In modern football, the difference is no longer just who scores —it is who consistently moves the game toward danger, and who quietly moves it away.