Value bet: what the edge actually measures

What is a value bet?

A value bet describes a situation where the estimated probability of an outcome exceeds the probability the offered odd would require to be balanced. It is a comparison between two numbers: a price, and an estimate that does not come from that price.

The concept is theoretical, and everyday language often betrays it. "Value bet" does not mean "bet to place", nor "bet that will win". It describes a gap calculated from assumptions — and a gap is never better than the assumptions that produce it.

The fundamental condition: two sources, never one

A value calculation requires an odd and a probability estimated independently of that odd. Without this independence, the calculation has no content.

The classic mistake is to use the bookmaker's implied probability as the estimate. The result is mechanically negative, by the amount of the margin: comparing a price with itself reveals nothing but the overround. Even a margin-free odd does not constitute an independent estimate — it is still extracted from market prices, and it inherits all their errors.

A usable estimated probability therefore comes from elsewhere: a statistical model, a consensus of several sources, a qualified external data point, or a personal analysis acknowledged as such.

The formulas

Two formulas are enough, and they say the same thing from two angles.

theoretical edge = estimated probability × decimal odd − 1

expected value = theoretical edge × stake

With an odd of 2.20 and a probability estimated at 50%: 0.50 × 2.20 − 1 = +0.10, that is +10%. For a €10 stake, the theoretical expected value is +€1.

Another way of reading the same calculation: the fair odd corresponding to 50% is 2.00 (see fair odds). The offered odd, 2.20, is higher — hence a positive edge. The two approaches are equivalent; the second is often more telling, because it compares two prices rather than a price and a percentage.

Where does the estimated probability come from?

Four usual sources, each with its own blind spots.

A statistical model. The most rigorous on paper, provided it is calibrated and evaluated out of sample. A model that has never been tested on data it has not seen has no probative value.

A personal estimate. Legitimate, but rarely calibrated. The difficulty is not naming the favorite — the market does that too — but quantifying a gap to within a few points, something intuition does not do well.

A consensus of several sources. Solid against isolated errors, but correlated with the market: if every source repeats the same prices, the consensus is no longer independent of the price it is meant to evaluate.

External data. Lineups, form, playing conditions, injuries: useful if it is more recent than what the market has priced in. That is increasingly rarely the case — the speed at which information is priced in is exactly what odds movements measure.

Why can an estimated probability be wrong?

Five mechanisms, all discreet in the resulting figure.

• Overfitting. A model calibrated on too few matches reproduces the noise of its sample and looks excellent on its own history.
• Incomplete data. A missing variable — a late withdrawal, a stake in the standings — shifts the estimate without making it visibly absurd.
• Poor calibration. A source that announces 60% in situations where the outcome occurs 45% of the time produces positive edges in a series, all of them wrong.
• Selection bias. Keeping only the matches where the model "got it right" manufactures a performance that does not exist.
• Stale information. The market prices continuously; an estimate produced the day before can lag behind a price adjusted since.

None of these problems shows up in the final figure. A +10% edge from a solid estimate and a +10% edge from an overfitted model are written exactly the same way.

Is a margin-free odd a value bet?

No. The question comes up often enough to deserve its own section.

A margin-free odd describes what the market estimates, once the overround has been redistributed. It is almost always higher than the displayed odd — since the margin has been removed from it — but that increase does not constitute an edge. It only measures the margin.

For there to be theoretical value, an estimate is needed that contradicts the market, not a restatement of the market. The distinction is the same as the one between margin-free odds and a model's fair odd, covered on fair odds.

Value bet or winning bet?

The two concepts are independent, and confusing them is the most costly mistake on this subject.

A value bet can lose. A +10% edge on an outcome estimated at 50% means it loses half the time: that is what the calculation predicts, not an accident.

A winning bet can have been placed at a poor price. An outcome at 2.00 taken at 1.70 remains a poor price, even if it happens. The result says nothing about the quality of the price — this is precisely why some people look at closing line value, which compares two prices without ever consulting the match result.

Finally, neither of these concepts makes a bet risk-free. They describe theoretical expected values over a large number of repetitions, within a framework where each individual bet remains uncertain and losses are possible. Responsible gambling is not a stylistic clause in this context: it is the direct consequence of the fact that a theoretical positive expected value protects against no actual loss.

OddScore's position

OddScore helps you understand the market price. It does not claim to guarantee that an odd constitutes a value bet.

The platform converts odds into probabilities, removes each operator's margin, estimates a reference price and tracks how it evolves. It does not produce a sporting probability independent of the market, and therefore does not have the ingredients needed to identify a value bet — no wording on the site will ever do so.

Dig into the market

Odds movements are only part of the story. Here are the next topics to read.

Frequently asked questions

Sources & methodology