Expected Threat (xT) Explained: Beyond xG

Arsenal's opening goal in their 3-1 win over Burnley in December 2018 was a beauty. Mesut Özil sliced the defence open with a through ball to Sead Kolašinac, who cut it back across the face of goal for Pierre-Emerick Aubameyang to finish. The assist went to Kolašinac. The goal went to Aubameyang. Özil got nothing.

Expected Threat, or xT, tells a different story. By its reckoning, Özil's pass lifted Arsenal's chance of scoring within the next five actions from 7.7% to 15.8%. Kolašinac's cutback added just 1.3 percentage points on top. Özil did 86% of the work, and the stat sheet never noticed.

That is the appeal of xT. Goals, assists and even xG only count the last step of a move. xT credits the passes and carries that made the last step possible, a bit like judging a chess game by more than the final capture. Here is how it works, where it came from and how a bettor can put it to use.

Expected Threat puts a number on how much every pass, carry and dribble changes a team's chance of scoring. Expected Goals (xG) only rates shots. xT rates everything else on the pitch too, including the moves that never end in a shot at all.

The idea is beautifully simple. Chop the pitch into a grid, and give each zone a threat value: how often possession there leads to a goal within the next few actions. Say the ball in your own box is worth around 0.01, a 1% chance. Level with the penalty spot it might be worth something like 0.35. Move the ball from one zone to the other and the difference is the credit that player earns.

So a pass from your own half to the edge of the opposition box does more than gain territory. It measurably raises your chance of scoring. Key difference: xG values the shot, xT values the build-up.

Under the bonnet, xT asks a simple question from every spot on the pitch: shoot, or move the ball on? The model weighs the chance of scoring if you shoot against the value of wherever you are likely to move it next.

The calculation runs in rounds until the numbers settle, usually after four or five. Each round adds one more action to the chain, so after five rounds the figure for a zone is the probability of scoring within the next five actions from there. In symbols:

The maths behind xT is older than its name. In 2011, Sarah Rudd presented a paper at the New England Symposium on Statistics in Sports at Harvard, using Markov chains to value possession across a handful of pitch states (box, wing, midfield, goal, lost). Her model rated a young Jordan Henderson, then at Sunderland, as a top-25 Premier League player, while prolific striker Darren Bent ranked poorly. It looked daft at the time. It looks far less daft now.

Karun Singh then brought the idea to a wider audience. His December 2018 blog post gave it the name "Expected Threat", added interactive visualisations and laid out the method in enough detail for anyone to copy. The name stuck, and his post is still one of the most-cited references on the subject.

The usual set-up carves the pitch into a 16x12 grid, 192 zones in all. Every zone carries four numbers drawn from historical data: how often players move the ball from there by passing or dribbling, how often they shoot (the two add up to 100%), where the ball tends to go next, and how often a shot from that zone ends up in the net.

The model starts with every zone at zero and builds up in rounds. Round one is effectively an xG model, because only shooting counts. Round two adds "move, then shoot". Round three adds "move, move, shoot", and so on. By round four or five the values stop changing much, and each zone's number is its chance of producing a goal within five actions.

Why stop at five? Longer sequences get increasingly unpredictable, and the extra rounds add very little. It is a neat trade-off, and it keeps the numbers easy to interpret.

Scoring an individual pass or carry could hardly be easier: take the threat value where it ended and subtract the value where it started. Move the ball from a zone worth 0.05 into one worth 0.15 and that action created 0.10 xT, ten percentage points of extra scoring chance within five actions.

Add those gains up across a match or a season and you have a measure of how much threat a player or a team creates, whether or not any of it ends in a shot.

The idea behind xT was spelled out in basketball analytics by Cervone and colleagues: "Most advanced metrics remain based on simple tallies relating to the terminal states of possessions... This is akin to analyzing a chess match based only on the move that resulted in checkmate, leaving unexplored the possibility that the key move occurred several turns before."

Football has exactly the same blind spot. Goals, assists and xG all sit at the end of a move. They miss the progressive pass that broke a defensive line, the carry that dragged defenders out of position and the switch of play to an overloaded flank. And since most on-ball actions in a match are passes and carries rather than shots, that is a lot of football going uncredited.

Expected Goals revolutionised how we judge shot quality, but a line-breaking pass, a dribble that opens space or a carry out from deep earns nothing in xG. xT and xG measure different skills, so a player can be superb at creating threat without taking many shots.

Go back to Özil and Kolašinac. Özil's through ball took the threat from 0.077 to 0.158, worth 0.081 xT and 86% of the credit. Kolašinac's cutback took it from 0.158 to 0.171, worth 0.013 xT and 14%. The assist column gave all of it to Kolašinac, yet Özil's contribution was about six times bigger. That matters for scouting, for contract talks and, as we will see, for anyone trying to find an edge in the markets.

The Henderson test

The best advert for possession-value thinking is still Sarah Rudd's 2011 work. Her model picked out Jordan Henderson as a top-25 Premier League player while rating Darren Bent poorly. At the time that sounded absurd: Bent was a proven goalscorer, Henderson an unremarkable midfielder. Henderson then joined Liverpool for a reported £16-20 million, and Bent joined Aston Villa for up to £24 million.

Years later Henderson was lifting the Champions League, the Premier League, the FA Cup and the League Cup as Liverpool captain. The analysts had spotted something that goals and assists could not show.

The unsung full-back

The same logic explains why full-backs and deep midfielders so often look better in xT than in the traditional columns. They rarely score or assist, but they move the ball from deep into dangerous territory again and again. That work never shows up on a highlights reel, yet it is exactly what xT is built to count.

Reading a team

Per-team xT maps show where a side creates its danger: down the left, through the middle, from deep carries or from wide deliveries. Flip the view and you can see where an opponent is hurt most, then combine it with pass destinations to find the passages of play to defend. Plot xT through a match and you also see who is really on top. A team can dominate the threat timeline without registering many shots, simply because its build-up keeps getting the ball into dangerous areas.

Judging players

xT rewards the players the usual columns overlook: full-backs, midfielders and anyone else who moves the ball forward without finishing the move. It also splits neatly into passing and carrying. Kevin De Bruyne is the classic passer, Eden Hazard the classic carrier, and Lionel Messi the player who could do both. Compare the action a player chose with the best one available and you start to see decision-making too.

Finding the next signing

For recruitment, the question is fit. A possession-based side wants a different xT profile from a counter-attacking one. And because xT surfaces players who drive build-up without big goal and assist tallies, scouts can use it to spot talent in smaller leagues before the bigger clubs catch on.

No metric is perfect, and xT's simplicity is both its charm and its weakness. The researchers behind the VAEP framework at KU Leuven put it bluntly: xT's location-based approach is not the most accurate, because the threat lies in how the ball is moved rather than where it is.

No sense of pressure or defenders

A pass into the box against a compact low block scores the same as one against a stretched defence. Without tracking data, xT only sees the ball, not the defenders around it, so a zone that looks dangerous on the map can be packed with bodies. The same action in a different context can be vastly harder.

No sense of risk

A risky pass into the box is worth the same as a safe one if both arrive. Basic xT also ignores failures: an intercepted through ball scores zero, even if it was exactly the right ball to try. A fast break against a disorganised defence should be worth more than a patient build-up that ends in the same spot, and xT cannot tell them apart.

No sense of the game state

Scoreline, minute and tactical situation are not in the model. A team 3-0 down in the 85th minute may rack up plenty of xT while the opposition sit deep, but that is not genuine threat. Treat the numbers as a guide to process, not a verdict.

Those gaps have inspired a family of smarter possession-value models. VAEP, from KU Leuven's DTAI lab, uses machine learning on rich game-state features and values both scoring and conceding. StatsBomb's OBV builds on its xG model and uses 360 data to see where the players around the ball are. American Soccer Analysis has its own Goals Added (g+) for MLS.

They all chase the same goal as xT, valuing every action rather than just the last one, but each adds the context that basic xT lacks. The table shows where each one fits.

xT will not hand you winners, but it is a sharp lens for working out who is better than the results say. Used alongside a proper football betting model, it points to a few practical angles.

Teams whose results lag their process

A side that keeps creating dangerous positions but converts them poorly is a good bet to regress upwards. xT measures underlying performance independently of finishing luck, so it helps you separate a genuinely poor team from an unlucky one. The same thinking can help with totals, which is where our guide to Over/Under goal lines comes in.

Player and assist markets

Midfielders who create plenty of threat with their passing may be underrated in assist markets. Their teammates' finishing can hide how good the underlying process is, which is exactly where value tends to hide.

Tactical mismatches and live betting

Hold a team's danger zones up against the opponent's weak spots and you can find a matchup edge. If Team A creates most of its threat from wide areas and Team B concedes heavily from crosses, that tells you something. In play, a side that is dominating the xT timeline while trailing on the scoreboard can be a genuine live betting opportunity.

The next wave of metrics is already taking shape. Tracking data will let analysts account for defensive positioning, reward off-the-ball movement and measure space created, not just ground gained. Dynamic Expected Threat (DxT) and similar models aim to fold in defensive pressure, game state and transition speed, the context basic xT lacks.

Expect more risk-and-reward thinking too, with models that weigh the chance of scoring against the chance of conceding, especially for high-risk line-breaking passes. As possession-value models become standard at StatsBomb and many clubs, comparing methods and benchmarks gets easier, and the public conversation gets smarter with it.

Expected Threat is one of the most useful ideas in football analytics since xG. It will not tell you who wins on Saturday, but it shows you the moves before checkmate: the passes, carries and dribbles that create chances before anyone pulls the trigger. It is not perfect, but a simple model that gets you most of the way, quickly, beats a complicated one you never use.

• Back process over results. A team creating plenty of xT but losing games is more likely to turn than one winning on lucky finishes.
• Hunt the hidden creators. Passers and carriers who build the move rarely get assist credit, so look for them in player and chance-creation markets.
• Match danger zones to weak spots. Where a side creates its threat against where its opponent leaks is a cleaner angle than raw form.
• Respect the context. Pressure, game state and risk are not in basic xT, so use it alongside xG and your own eyes.
• Look for the live mismatch. A team dominating the threat timeline while trailing is the sort of in-play situation worth a closer look.
• Keep stakes sensible. No metric removes variance, so stake what you can afford and never chase losses.

The assist tells you who passed. The goal tells you who scored. xT helps show who really created the chance.

• How to Use xG (Expected Goals) for Smarter Football Betting
• How to Build a Sports Betting Model: Complete Guide
• Home Advantage vs Team Quality: Advanced Football Betting Guide
• Minute Betting Markets: When Goals Happen

18+. Please gamble responsibly. For free, confidential support visit BeGambleAware.org.