A final score is an answer. The trace is the proof.
The engine does not predict a scoreline. It resolves a match as a sequence of chances on a ninety-minute timeline, with the referee and the manager acting inside it, then runs it thousands of times and reports the distribution: who wins, how often it is a draw, the scoreline spread, the goal margin, the total. With the events switched off it reduces to the same goal model the betting markets use, so it is calibrated against them first. Football is a low-scoring, three-outcome game, so a strong favourite still carries a real chance of dropping points, and every result comes with the trace of why.
Not one scoreline. A distribution, calibrated to the market.
Goals come from chances, so the chance is the unit, placed on a timeline and resolved thousands of times. The answer is never a single scoreline; it is the full distribution, and its three pieces map directly onto the three objects the betting markets price. This is a composite favourite at home against a mid-table side, ten thousand runs.
The margin is the Asian Handicap, the total is the over-under, and the win-draw-loss is the 1X2, so the engine is calibrated against the sharpest football forecasters there are before any events are layered on. And it refuses to manufacture certainty a low-scoring sport does not contain: a real fifty-six beats a false ninety. The answer to a match is a shape, and the shape is priced against the market.
Illustrative engine read on the real chance-level, multi-run distribution (win-draw-loss, scoreline spread, goal margin as the Asian Handicap, total as the over-under, the xG line), market-calibrated. Composite teams, demonstration figures.
Two people move a football match. The engine models both, honestly.
In other sports the official is a foul rate. In football the referee swings matches: a sending-off removes a player for the rest of the game, a penalty is a near-goal, added time extends the window. So he is a first-class actor with a tendency vector, shrunk to the league mean by how little he is sampled. The manager, by contrast, adds no base multiplier, because his system is already in the Team KR; his real edge is the bounded in-game layer that decides tight matches.
Run the same fixture under a strict referee and a lenient one and the distribution moves, because a chippy derby is officiating-sensitive while a mismatch is not, and that sensitivity is itself the insight. The manager will not inflate a bad team into a good one, but he will win or lose the tight one at the hour mark. The engine gives each actor exactly the weight the evidence supports, no more. The referee and the manager are real match-swingers, priced honestly rather than assumed away or overblown.
Illustrative engine read on the real referee layer (the RIF tendency vector, the sending-off swing, sample shrinkage, scenario override) and the manager layer (no base multiplier, the bounded quality-gated in-game menu). Composite match, demonstration figures.
Not just who wins. Exactly what won it.
A win probability with no explanation is a number to be taken on faith. After the runs, the engine attributes the result: how much the scheme battle moved the margin, which players swung the win probability, which archetype-versus-style conflicts drove it, and which discrete events turned matches. The distribution is the answer; this is the proof.
Every one of these is recoverable because the match is resolved event by event on a timeline, so the engine can say not only that the favourite wins fifty-six percent of the time but exactly which matchup, which player, and which events produced that number. A black box hands you the probability and hides the reasoning. This hands you both. The distribution tells you what happens; the trace tells you why, and only the second one you can act on.
Illustrative engine read on the real attribution structure (class attribution, swing players via like-for-like replacement, scheme conflicts, decisive referee-mediated events). Composite match, demonstration figures.
A match is not a number to be guessed; it is a sequence of chances on a timeline, with a referee who swings it and a manager who shapes it in-game, run enough times to become a distribution. That distribution is calibrated against the markets that price football most sharply, and it refuses to manufacture the certainty a three-outcome, low-scoring sport never contains, so the favourite carries a real chance of dropping points. And because every run is resolved event by event, the engine can hand back not just the probability but the proof: the scheme battle, the swing player, the matchup, the events that turned it. A score you cannot open is a guess with a decimal point. The trace is what makes it intelligence.
Run the match. Read the proof.
Simulation resolves a match chance by chance, models the referee and the manager as the actors they are, and hands back a market-calibrated distribution with the full trace of why.