Odds vs Model Probability
Odds and model probability are not the same number: a market price converts to "implied probability" via 1/odds but carries the operator's margin (vig), while a model probability comes from independently processed data. Sound analysis reads the market as an information-bearing signal and an input — never as certainty. This page explains implied probability, overround, why the two numbers diverge and what line movement tells you. It is analytics, not betting advice.
Implied probability: 1 / odds
Every decimal price carries an implicit probability claim: implied probability = 1 / odds. A price of 2.00 implies 50%, 1.50 implies 66.7%, 4.00 implies 25%. That conversion is the first step in putting a market price and a model output into the same language.
The raw conversion is incomplete, though: a market price contains an operating margin, so the implied probabilities of the three outcomes sum to more than 100. Serious comparison strips the margin first; the underlying vocabulary is collected in the glossary and unpacked step by step here.
Vig / overround: the margin inside the price
In a typical match the 1X2 implied probabilities sum to roughly 105-108%; the excess over 100 is the overround (margin, vig). Example: prices of 2.10 / 3.40 / 3.60 imply 47.6% + 29.4% + 27.8% = 104.8%. Since real-world outcome probabilities must sum to 100, the extra 4.8 points are the cost layer of the price.
Redistributing the margin proportionally ("no-vig" normalisation) recovers an approximation of the market's true probability opinion. 11Stat uses this normalised baseline in its comparisons — measuring a model against raw prices is systematically misleading.
Why a model probability diverges from the market
The two numbers come from two different processes. The market price is the composite of money flow, news flow and pricing policy. The model probability is produced by running data — xG, form, squads, league context — through Poisson/Dixon-Coles-based engines, documented openly on the methodology page.
Divergence has three typical causes: the model is processing information the market has not priced yet; the market is pricing information the model cannot see (an injury rumour, a leaked lineup); or one of the two is simply wrong. Divergence alone is not "an opportunity" — the first check is the model's calibration: divergence from a model whose numbers cannot be trusted is noise, not signal.
What line movement tells you
Prices do not sit still between opening and kickoff: new information (lineups, injuries, weather) and money flow keep updating them. The opening price is a first estimate made with relatively little information; the closing price is the market's final estimate after absorbing everything — empirically the most accurate single forecast available.
Direction of movement is therefore an analytical signal: a steadily shortening price means the market is revising that outcome's probability upward. One of the most honest ways to measure a model's long-run quality is to ask where its estimates stood relative to the close — the topic of the CLV page.
Odds are an input, not a promise: how 11Stat uses them
11Stat uses market prices for three analytical purposes: (1) showing the no-vig market probability side by side with the model's output, (2) reporting model-market divergence and backtesting whether it has historically carried meaning, and (3) tracking model quality against the closing price (CLV). The full outcome distribution behind the most likely score page follows the same principle.
At no point is the price treated as an instruction about what to do; it is a reference line that measures how the model aligns with the outside world. 11Stat gives no betting advice — it processes odds strictly as data.
Common misconceptions
"Short odds mean near-certainty." False. A 1.30 price is roughly a 73-75% probability once the margin is stripped: the favourite dropping points about one match in four is expected behaviour, not an upset of the system.
"The price moved, so the result is settled." Movement is a probability revision, not a settlement; a probability revised from 60% to 70% still goes the other way three matches in ten. "If the model disagrees with the market, the model is right." That is a hypothesis, testable only through calibration and long-run backtests; a single match proves nothing. Football is a high-variance sport, and no number erases that variance.
Frequently Asked Questions
Do short odds mean the favourite will win for sure?
No, they do not. Short odds only mean the market assigns a high probability; even a 1.30 price loses roughly one time in four. "For sure" does not exist in a probabilistic sport, and content that uses that language should not be trusted.
Is a heavily backed favourite a sure bet?
No — a sure bet does not exist. A shortening price means the market has revised the probability upward, not that the outcome is settled: even a 70% event fails three times in ten. Certainty language is a marketing device, not a statistical statement.
If the model and the market disagree, which one should I trust?
Both are hypotheses. The sound approach looks at process, not one match: if the model's calibration report holds up and its divergences have historically stood well against the closing price, it earns weight; otherwise the divergence is noise. 11Stat publishes both tests openly.
Why does the overround (vig) exist?
It is the pricing side's operating margin: it pushes the sum of implied probabilities above 100, adding a cost layer to the price. On the analysis side, no model-versus-market comparison should be made before this margin is stripped via no-vig normalisation.
Does 11Stat give odds tips?
No. 11Stat provides no tips, slips or betting advice. Odds enter the system purely as data: the no-vig market probability is displayed next to the model output, and model quality is reported against the closing price (CLV). Decisions and responsibility always remain with the reader.
Why is the closing line considered more accurate than the opening line?
Because by the close, lineups, injuries and the full information flow have been priced in, whereas the opener is a first estimate made with limited information. Empirically the closing price is the strongest known single predictor of match probabilities — still not a certainty, just the least-wrong forecast.