Expected Threat (xT) — The Football Analytics Metric That Values Every Touch

June 24, 2026 · 10 min read

Expected goals tells you how likely a shot is to become a goal. But what about the through-ball that created the shot? The dribble that broke the defensive line? Expected Threat (xT) assigns value to every ball movement on the pitch — and it is changing how clubs evaluate players and predict match outcomes.

The Blind Spot in Expected Goals

Expected goals (xG) transformed football analysis. For the first time, fans and analysts could look beyond the final scoreline and ask: did a team actually create enough chances to win? A team that generates 2.5 xG but scores once is unlucky; a team that generates 0.3 xG and wins 1-0 rode its luck.

But xG has a fundamental limitation. It only values shots. Every other action on the pitch — every progressive pass, every line-breaking dribble, every switch of play that pulls a defense apart — registers as zero in a pure xG model. Consider Mesut Özil’s famous through-ball to Olivier Giroud: the pass splits two defenders and puts Giroud one-on-one with the keeper. Giroud’s shot gets credited with high xG. Özil’s pass? Nothing.

This is the blind spot. Football is a game of territory. Moving the ball from your own half into the opposition’s penalty area is valuable, even if no shot results from that particular possession. Expected Threat was designed to fill this gap.

What Is Expected Threat?

Expected Threat (xT) is a pitch-control model created by data scientist Karun Singh. Instead of valuing only shots, xT assigns a threat value to every zone on the football pitch. That value represents the probability that a possession starting in that zone will result in a goal within the next few actions.

The model divides the pitch into a grid — typically 16 columns by 12 rows, creating 192 zones. Each zone gets a value between 0 and 1. Zones near the corner flags sit close to 0. Zones inside the six-yard box approach 0.9 or higher. Every other location falls somewhere on that spectrum.

The value of any single action — a pass, a dribble, a carry — is then calculated as the difference between where the ball started and where it ended up:

xT gained = xT(end zone) – xT(start zone)

If a midfielder receives the ball in a zone valued at 0.04 and plays a progressive pass into a zone valued at 0.12, that pass generated 0.08 xT. It moved the ball into a position that is twice as dangerous as where it started. Over a full match, summing every player’s xT contributions reveals who is actually driving a team’s attacking play.

The Math Behind It: Markov Chains on a Football Pitch

Under the hood, xT uses a mathematical framework called a Markov chain. A Markov chain models a system where the next state depends only on the current state, not on how you got there. On a football pitch, the “state” is which zone the ball is in, and the “transition” is a pass or dribble that moves it to another zone.

For each zone, the model calculates three probabilities from historical match data:

• Shoot probability (s): How often a player shoots when receiving the ball in this zone
• Move probability (m): How often a player passes or dribbles instead of shooting (m + s = 100%)
• Goal probability (g): If a player shoots from this zone, the chance of scoring — essentially a simple xG model per zone

The core equation ties these together:

xT(x,y) = (s × g) + (m × Σ T(x,y)→(z,w) × xT(z,w))

In plain language: the threat of a zone equals the chance of scoring by shooting from there, plus the chance of moving the ball somewhere else multiplied by the threat of that destination. The equation is circular — xT at every zone depends on xT at every other zone — so it is solved through iteration. Start with all zones at zero, recalculate, and repeat until the values stabilize.

After one iteration, the model essentially reproduces xG. After two iterations, it captures “move, then shoot.” After five iterations, the values represent the probability of scoring within the next five actions from any zone. This multi-step lookahead is what makes xT powerful: it captures the cascading value of ball progression.

xT vs xG: A Side-by-Side Comparison

The two metrics are complementary, not competing. xG tells you about the quality of chances a team creates. xT tells you about the quality of play that leads to those chances. A team with high xG but low xT is relying on individual finishing brilliance. A team with high xG and high xT is building dangerous attacks systematically.

How Clubs Use xT in Practice

Scouting and Recruitment

When a club scouts a central midfielder, traditional stats show pass completion rate and assists. xT shows something deeper: does this player consistently move the ball into more dangerous zones? A midfielder with 92% pass completion who mostly plays sideways passes will have low xT. A midfielder with 85% completion who regularly plays progressive passes into the final third will have high xT. The second player is more valuable, even though his raw accuracy is lower.

Evaluating Build-Up Play

Karun Singh’s original analysis broke down a single Arsenal goal into its component passes. Özil’s through-ball earned 86% of the xT credit for the move, while Sead Kolašinac’s earlier cross earned 14%. This kind of attribution is impossible with xG alone, which would credit 100% to the shot.

Opponent Analysis

By mapping an opponent’s xT creation across the pitch, analysts can identify where the danger comes from. Does the opponent create most of their threat through the left wing? Through central through-balls? From set-piece routines? xT heat maps reveal attacking patterns that raw possession stats miss entirely.

Counter-Attack Detection

Tracking xT changes during a possession sequence reveals how quickly a team transitions from defense to attack. A possession that starts at 0.01 xT (own penalty area) and reaches 0.35 xT (opposition box) in three passes is a devastating counter-attack. xT makes these sequences measurable and comparable across matches.

What xT Means for Football Predictions

For anyone making football predictions — whether in a FanPick group or building a statistical model — xT adds a dimension that xG alone cannot provide.

Traditional prediction models rely heavily on xG, shots on target, and possession. These are useful but incomplete. Two teams can generate identical xG totals while playing completely different styles of football. Team A creates three high-quality chances from fast counter-attacks. Team B creates ten half-chances from patient build-up. xG says they are equal. xT reveals that Team B’s build-up is more consistently dangerous — they are sustaining pressure in high-threat zones, even if the individual chances are lower quality.

This distinction matters for predictions because sustained pressure correlates with future performance. A team that consistently generates high xT is more likely to maintain its scoring rate over a season than a team relying on sporadic high-xG chances. xT acts as a leading indicator: if a team’s xT is rising but their xG is flat, goals are likely coming.

The Limitations of xT

No metric is perfect, and xT has real constraints worth understanding:

• Data hungry: xT requires event-level tracking data (every pass, dribble, and shot with precise pitch coordinates). This data is expensive and not publicly available for most leagues.
• Grid resolution matters: A 16×12 grid works well for Premier League data with thousands of matches. Smaller leagues with fewer matches produce noisier zone values.
• No defensive value: xT only measures on-ball actions. A defender who positions himself perfectly to prevent a pass from ever being attempted generates zero xT. Defensive contributions remain hard to quantify.
• Context-free zones: The model assigns the same threat value to a zone regardless of how the ball arrived there. A zone valued at 0.15 is treated identically whether the ball came from a quick counter-attack or a slow build-up under pressure.
• Not a standalone predictor: xT works best as one input in an ensemble model, not as a prediction tool on its own. Combine it with xG, Elo ratings, and team form for the most accurate forecasts.

xT in the Broader Analytics Ecosystem

xT sits alongside a family of metrics that have expanded football analysis beyond goals and assists. Expected assists (xA) values the quality of a pass that leads to a shot. xG Chain credits every player who touched the ball during a possession that ended in a shot. Progressive passes and carries measure how often a player moves the ball significantly closer to goal. PPDA (Passes Per Defensive Action) quantifies pressing intensity.

Together, these metrics paint a richer picture than any single statistic. Data providers like StatsBomb, Opta (now Stats Perform), and Wyscout have made this data increasingly accessible, though the most granular tracking data still requires paid subscriptions. For prediction enthusiasts, the key insight is that ball progression metrics like xT are more stable match-to-match than finishing metrics, making them better predictors of future performance.

The clubs that consistently outperform their spending — think Brentford, Brighton, or Atalanta — are often the ones that identified players with strong underlying progression metrics before the market caught up. xT is one of the tools that makes those identifications possible.

Key Takeaways

• xT fills xG’s biggest gap: It values every on-ball action, not just shots, by assigning threat values to all 192 zones on the pitch.
• The math is elegant but accessible: A Markov chain model with iterative convergence — five iterations capture the probability of scoring within the next five actions.
• xT reveals hidden value: Players who consistently move the ball into high-threat zones generate more xT than flashy dribblers who lose possession in safe areas.
• Predictive power: Teams with rising xT but flat xG are likely to improve — xT is a leading indicator of future goals.
• Use it as one piece, not the whole puzzle: xT works best alongside xG, Elo ratings, and form data in an ensemble prediction model.