Recursive Expected Threat (xT) Model

Expected Threat (xT) is a zone-based model that assigns a goal-scoring probability to every location on the pitch by recursively computing: from this zone, what is the probability of scoring within the next N actions (typically 5)? Unlike xG (which only values shots) or EPV (which requires tracking data), xT works entirely from event data and values every on-ball action — passes, carries, dribbles — by the change in xT they produce. The pitch is divided into discrete zones (typically 150 zones of 8×8 meters), and the model iterates backward: first, what's the chance of scoring from a shot in each zone? Then, what's the chance of moving the ball to a higher-threat zone and scoring from there?

(1) Divide the pitch into N zones (8×8m grid = ~150 zones). (2) First iteration: for each zone, compute P(shot) × P(goal|shot) — the base threat from shooting. (3) Subsequent iterations: add the movement term — for each zone XY, sum over all destination zones ZW: P(move from XY to ZW) × xT(ZW). This captures the value of being able to pass or carry to higher-threat zones. (4) Iterate 5-6 times until convergence. (5) Per-action xT: for any action (pass, carry), compute xT(end zone) - xT(start zone). Positive = the action increased threat; negative = it decreased threat.

The resulting pitch map shows high xT near the goal (obviously), but also meaningfully positive xT in buildup zones where teams can progress toward goal. The key insight: xT values every action, not just shots, making it the most accessible all-action valuation framework for event data.

1. Confusing xT with xG: xG values shots only. xT values every action. They are complementary, not competing.
2. Confusing xT with EPV: EPV is continuous, uses tracking data, and models possession probability. xT is discrete, uses event data, and models threat from zones. xT is simpler and more accessible; EPV is more accurate.
3. Not including shots in xT: The original xT formulation sometimes excludes shot-taking. Including shot xT (shot xG minus zone xT = how well the player maximized their opportunity) gives a more complete picture.