Live Win Probability

What live win probability actually is

A pre-match analysis starts ageing the moment the whistle blows. Live win probability takes the pre-match model's output as a starting point (a prior) and updates it with every new piece of information from the pitch: has the score changed, how many minutes remain, who is applying pressure, are both sides still at eleven players?

The output is three numbers — home win, draw, away win — that sum to 100% at every instant. This trio is not a verdict but a snapshot of the current state: a side shown at 80% in the 70th minute can still lose; the number only reports the balance of evidence so far. The underlying vocabulary lives in the glossary.

The model's inputs: score, minute, live xG, momentum

11Stat's in-play signal engine blends four main inputs:

• Current score and time remaining: the dominant pair. The same 1-0 is fragile in the 20th minute and nearly decisive in the 85th.
• Live xG: the chance quality accumulated during the match, quantifying pressure the scoreboard hides. The concept is covered in the xG guide.
• Momentum: the density of shots, corners and dangerous attacks over the last 10-15 minutes — which goal the game is flowing towards.
• Structural shocks: red cards, penalties and forced substitutions, events that shift the whole distribution in one step.

Anchoring all of this is the pre-match 12-engine consensus: a strong favourite that concedes early does not lose its standing entirely, because the prior still carries weight.

Why a single goal moves the number so hard

Because football is low-scoring, one goal is an enormous piece of information relative to how many goals are still expected in the remaining minutes. Mathematically the model asks: "Is the expected goal production in the time left enough to overturn this margin?" That is why a 30th-minute goal moves the needle far less than the identical goal in the 85th.

An example: in a balanced match, a 1-0 scored in the 60th minute typically lifts the leading side from around 45% to the 70-75% band; the same goal in the 88th pushes it above 90%. As time drains, every event gets more "expensive" — a kind of time-value decay, and the reason live probability charts steepen towards the end.

Momentum and live xG: when pressure is signal, when it is noise

Everyone knows the story of the team that dominates without creating. Live xG earns its keep exactly there: it measures genuine chance quality, not possession. A side holding 70% of the ball while producing 0.3 xG looks dominant to the eye but is nearly silent to the model.

Momentum, meanwhile, is a short-window signal: a burst of dangerous attacks over the last quarter-hour nudges the near-term scoring probability upward — but the effect is a small, fast-decaying correction. Pressure does not reliably convert into goals, which is why the model treats momentum as a bounded extra factor rather than a multiplier. A wave of pressure never means "a goal is coming"; it only means "the probability of a goal has risen somewhat".

How 11Stat's in-play signal engine works — and where it stops

The engine takes the probabilities produced by the pre-match 12-engine consensus as its prior, updates them several times a minute with live data, and passes every update through a calibration layer. The goal is to give viewers an educational probability stream: following the story of a match in numbers and seeing what the model reacts to, and why.

The limits are equally explicit: data feeds can lag by seconds, rare events such as penalties and red cards cannot be anticipated, stoppage time is uncertain, and single-match variance is always bigger than the model. Live probability is therefore a reading instrument, not a decision instrument. 11Stat publishes the live stream for analysis and education; none of its output is a prompt to wager on anything.