What Is a Value Bet? We Tested the Textbook Method on 455 Matches
A value bet pays more than the true chance deserves. We estimated the true chance from four venues' prices, bet every outlier, and learned why the market is so hard to beat.
Last reviewed November 3, 2026
Key takeaways
- >Value = odds × true probability − 1; it is about the price, not the result
- >A margin-free consensus of four venues was within about 2 points of reality in every probability range (455 matches)
- >Backing every price ≥3% above that consensus: 334 bets, −8.4%, within luck
- >Outliers ≥5% above consensus lost beyond luck: the out-of-line venue is usually the one that is wrong
- >Two thirds of the apparent value sat on outsiders at 5.00 or longer
A value bet is a bet whose odds are longer than the true chance of the outcome deserves. It is the only kind of bet that makes money in the long run, and every serious betting method is, one way or another, a way of finding them. The hard part is the word "true". We tested the most common way to estimate it on our own data, and the result is more interesting than the textbook suggests.
The idea
If an outcome has a 25% chance, its fair odds are 1 / 0.25 = 4.00. At odds of 4.40 you are being paid more than the risk deserves:
Value = odds × true probability − 1
4.40 × 0.25 − 1 = +10%. In the long run, every €10 on bets like that returns about €11. At 3.80 the same bet is −5%: you would lose slowly, however often it happened to win.
A value bet can lose, and usually several in a row. Value is about the price, not the result.
Where does "true probability" come from?
Nobody knows it. The common practical answer: the market itself. Take several venues' prices, remove each one's margin, and average them. That consensus is your estimate of the true chance. When one venue's price is clearly longer than that consensus, you have found a candidate value bet.
Is the consensus any good? Remarkably
We built that consensus for 455 finished matches priced at four venues (Toto, Unibet, Polymarket and Overtime) and compared its estimates with what happened:
Consensus chance | Selections | Priced at (average) | Actually happened |
|---|---|---|---|
Under 15% | 150 | 10.2% | 10.7% |
15–30% | 626 | 23.7% | 23.6% |
30–50% | 348 | 38.9% | 39.4% |
50–70% | 168 | 58.8% | 58.3% |
70% and over | 73 | 78.8% | 76.7% |
In every range, the market's de-margined estimate landed within about two percentage points of reality. That is the background against which every "edge" has to be measured: the combined market is very hard to beat.
What happened when we bet the value
We then took every selection where one venue's price was at least 3% above the consensus of the other three, and backed it at that venue, €10 flat:
Selections | Bets | Hit rate | Expected (consensus) | ROI | Luck band |
|---|---|---|---|---|---|
Value ≥ 3%, all four venues | 334 | 15.6% | 16.9% | −8.4% | ±26.7% |
Value ≥ 3%, Toto and Unibet only | 166 | 17.5% | 18.0% | 0.0% | ±37.8% |
Value ≥ 5%, all four venues | 212 | 9.9% | 15.2% | −31.9% | ±31.4% |
And by price:
Odds of the value bet | Bets | Hit rate | Expected | ROI |
|---|---|---|---|---|
2.50–5.00 | 104 | 26.9% | 27.2% | +1.0% |
5.00 and higher | 228 | 10.1% | 12.0% | −12.9% |
Three lessons from this:
- Most "value" by this method sits on outsiders. Two thirds of these bets were at odds of 5.00 or more. Venues disagree most, in percentage terms, on unlikely outcomes.
- None of it showed an edge. Everything except the ≥5% line stayed inside the luck band. The mid-priced value bets almost exactly matched the consensus (26.9% against 27.2%).
- The biggest outliers were the worst bets. Prices 5% or more above consensus lost beyond luck. When one venue is far out of line, it is more often that venue that is wrong (or about to move) than the rest of the market.
So how do people find real value?
- Better information than the market, before the market has it: team news, a model that is genuinely more accurate on a niche. Rare, and short-lived.
- Prices that have not caught up yet. A venue that is slow to move after news. Our data is collected once a day, so it cannot see these moments; they typically last minutes.
- Smaller and less-watched markets, where the consensus itself is thinner.
And for everyone: always take the best price available. Even when you have no edge, a better price reduces what you lose to the margin, and that is value too. See Same match, four prices.
How we measured this: Edge Terminal's odds history, the last 1X2 price recorded before kickoff at each of four venues, for 455 finished matches (14 August – 31 October 2026, nine competitions) priced at all four. Consensus: each venue's prices with the margin removed proportionally, averaged over the other three venues. Value bets between 3% and 30% above consensus; €10 flat. Luck band: ±2 standard errors of the per-bet return. Snapshot taken 3 November 2026. Nothing here is a tip.