Value Betting Step by Step: Estimate, Test, Size the Stake
Value is a gap between two probabilities. One of them is your own guess, which is where this page spends its time.
A value bet is a bet whose odds are higher than the true probability of the outcome, as MadMarket and Sportmarket both define it. In practice you compare two probabilities: your estimate of the outcome, and the 1 ÷ odds the price implies. Expected value per 1 staked is your probability times the odds, minus 1. At odds of 3.18, an estimate of 34% is worth +8.12% (hypothetical). The catch is the estimate: if it is three points too high, the same bet is worth -1.42%. This page covers the sum, where the probability comes from, Kelly stakes, and what an exchange commission and the rules of Orbit Exchange change on a betting exchange.
Reality check
Most bettors end behind. A positive expected value is an average over many bets, not a promise about the next one. Decide what you can afford to lose first, and read the responsible gambling page.
What counts as value
A good bet can lose and a bad bet can win. Value is a claim about the price, which is why it can be worked out before the match, from your estimate, and not read off the result.
MadMarket's guide gives the definition in its plainest form: "Value betting involves placing bets where the odds offered are higher than the true probability of an outcome." Sportmarket's FAQ says a value bet is the same thing in nearly the same words. An odds figure and a probability are different units, so the sum needs a conversion first, and the conversion is 1 ÷ odds. A converter for implied probability does it for any format. MadMarket's glossary then states the idea as one probability against another: "This is a bet with a higher true probability than the odds reflect."
AsianConnect's blog turns that into a test you can run: "If your predicted chance of something happening is higher than the bookmaker's implied probability, you have value." Two probabilities, one gap. (The word "edge" turns up in betting writing for the bettor's advantage in one place and the bookmaker's in another. Here, outside quotations, it means expected value per 1 staked.)
Value is not victory. AsianConnect's comparison page states the difference: "A bettor may correctly predict many matches while consistently accepting poor odds." Picking winners and finding prices that are too long for the chance are two separate skills, and everything below is about the second.
When your estimate is built from another book's prices, that book's margin has to come out first. The closing price and the margin are worked through on their own page, and this one starts after that step.
Expected value, with numbers
The expected value of a bet of 1 is your chance times the odds, minus 1.
Write it as a line: expected value per 1 staked = estimated chance × odds - 1. BetInAsia's calculator page writes the same idea in two terms, "Expected Value = (Probability of Win x Total Payout for Win) + (Probability of Loss x Total Payout for Loss)". Count the profit as the payout on a win and the lost stake as the payout on a loss, and the two forms give the same number.
The sign tells you what the price is worth against your estimate. MadMarket's guide reads a positive result as a value bet and a negative one as odds that are "not in your favour". A zero sits exactly on the line: the chance you believe equals the chance the price implies.
Expected value is an average. A Pinnacle article defines it as "how much we will win or lose on average in the long run if a bet like this was to play out many times". One bet does something else. It wins, or it loses.
Odds of 3.18, an estimate of 34%
Suppose a price of 3.18 and an estimate of 34%. The price implies 1 ÷ 3.18 = 31.45%. Expected value per 1 staked is 0.34 × 3.18 - 1 = +8.12%. At an estimate of 31.45% the figure is zero, which is the break-even chance of that price. All figures here are hypothetical.
| Estimated chance | Expected value per 1 staked | Full Kelly | Half Kelly | Quarter Kelly |
|---|---|---|---|---|
| 28% | -10.96% | 0.00% | 0.00% | 0.00% |
| 30% | -4.60% | 0.00% | 0.00% | 0.00% |
| 32% | +1.76% | 0.81% | 0.40% | 0.20% |
| 34% | +8.12% | 3.72% | 1.86% | 0.93% |
| 36% | +14.48% | 6.64% | 3.32% | 1.66% |
| 38% | +20.84% | 9.56% | 4.78% | 2.39% |
Read down the column, not across a single row: a few points of chance move the result a long way. Between 30% and 32% the sign flips. A table this steep is a warning about the input as much as a result.
Where the probability comes from
The price is on the screen. The chance is yours. Three places it can come from are named in the sources, and each is attributed here because none is a fact about your bet.
The first is another book's price. MadMarket's calculator tells you to check "Sharp" bookmaker prices and "remove the margin", and its guide says to use your calculations to "determine the implied probability of an event and establish the true price of a selection". The second is a benchmark. A Pinnacle odds screen says "many professional bettors use Pinnacle odds as a benchmark", which is Pinnacle describing itself, and it adds that a bettor may notice "a line has already moved at sharper books but remains available elsewhere". The third is a model of your own. AsianConnect's blog lists statistical models among the ways people look for value, and the work of building one is the subject of testing a model out of sample.
On where value is easier to find, the two sources give different answers. MadMarket's guide writes of sharp books that value opportunities there "are less frequent", and of soft books that "Their markets are often less efficient". This page does not settle it, and no venue is named as the place to look.
Now the weak point. A contributor to Pinnacle's resources writes that "knowing" the true probability of a complex system like a football match is "effectively impossible", and that if you believe you win 52% of your even-odds bets and win only 49%, "you are going to lose money over the long term". A Pinnacle article describes a margin of safety: set a threshold of expected value and take no price below it, so that the price sits above your own estimate of fair value. A single bet that clears the threshold does not make the guess any better.
I would trust a probability I can explain over one I can only feel.
When your estimate is off
Say you believe 34% and the true chance is 31%. At 3.18 the same bet is worth 0.31 × 3.18 - 1 = -1.42%, and Kelly stakes nothing. The bet that looked like value was a guess that was three points too high.
Run your own numbers
The calculator below opens on the numbers from this page, so its first result should match the example above. Change any of them. Add a commission if the price is on an exchange.
Expected value and Kelly stake calculator
Enter the odds, your estimated chance and your bankroll. Add a commission if the price is on an exchange.
Kelly sizing assumes your probability is right; it is an estimate, which is why a fraction of Kelly is often used.
Expected value is an average over many bets. One bet wins or loses.
This is arithmetic, not a prediction or a recommendation.
Example shown. Replace the numbers with your own.
The tool is arithmetic. It does not know whether your 34% is any good. This explains how value is measured. It does not tell you which bets to place.
Sizing a stake with Kelly
Once an expected value is positive, the next question is how much to stake. The Kelly formula is on the roadmap for professional bettors: f = (b × p - q) ÷ b, where b is the odds minus 1, p the chance of winning and q is 1 - p. MadMarket's glossary says the criterion "calculates the percentage of a player's bankroll to stake on bets with positive expected value", with long-term growth as the aim. That is an objective, stated by the source, and not a result you are owed.
Here the formula uses this page's numbers. At 3.18 and 34%, full Kelly is 3.72% of the bankroll. MadMarket's calculator says "If your estimated chance is lower than the break-even probability of the odds, no stake will be recommended", and the table agrees: every row at or below 30% stakes 0.00%.
- Full Kelly at 34% and 3.18
- 3.72% of the bankroll
- Half Kelly
- 1.86%, a stake of 74.50 on 4,000
- Quarter Kelly
- 0.93%, a stake of 37.25
- Full Kelly if the true chance is 31%
- 0.00%
Fractions are a practice, not a rule of this site. MadMarket's calculator proposes Full, Half or Quarter Kelly and writes "Use Half or Quarter Kelly to reduce bankroll volatility and protect against over-estimating your edge." A Pinnacle article writes "Experts like to use a fractional Kelly method and 30 percent Kelly is a common figure." The calculator above offers the three MadMarket fractions. None is recommended here.
Two limits come from contributors to Pinnacle's resources. One consequence of betting with the Kelly criterion is "a significant volatility in returns". And the usual formula is "a simplified form that works when there is only one bet at a time", so the line above is a starting point for a single bet, not for several open together. The stake is only as sound as the chance behind it.
On an exchange the commission comes first
At a book the price is set by the book, and its margin is already inside it. On an exchange the price is what other customers offer: backers and layers face each other. Orbit Exchange works that way, and how backing and laying fit together shows the mechanics.
The charge arrives afterwards. Orbit's betslip reads "A commission is charged on net profit from each market", and its tooltip describes the result as "The amount that you win minus the charge." Sportmarket says the same thing about exchanges generally, "You pay only on net winnings in that market, not on losses", and asks you to do something with it: "Factor this into expected value when comparing a sharp book price vs exchange."
The sum is short. Net odds = 1 + (odds - 1) × (1 - commission). Take an example rate of 3%, chosen for the sum and not a rate of Orbit's. Odds of 3.18 become 3.1146, so the break-even chance moves from 31.45% to 32.11%. At an estimate of 34% the expected value falls from +8.12% to +5.90%, and full Kelly from 3.72% to 2.79% (half Kelly, 1.39%). A charge that looks small on the screen takes a visible share of the expected value. The commission page carries the rest of the arithmetic, and comparing a net price with a book price compares them across venues.
Orbit's exchange rules add a second condition. In football markets they say "all unmatched bets are cancelled if the goal, penalty or red card is awarded by the referee", and that "a customer can request that an unmatched Exchange bet should not be cancelled when the market is turned in-play". The exchange also "reserves the right to suspend any market at any time at its sole discretion". An expected value attaches to money that is matched. An offer that sits unmatched earns nothing, however good the sum looked. How much a price can take is the place to read about that.
Price at a book, margin inside the price
You read the odds and work out the chance they imply.
- The book's margin is already in the number on the screen.
- One price to compare with your estimate.
Best for: a bettor who wants to read the book's margin straight from the odds.
Not ideal if: you want to see what you keep after a charge on winnings.
Price on an exchange, commission on net profit
You read the odds, then take the commission off the profit.
- The net price is lower than the one displayed.
- The break-even chance rises with the rate.
- Only matched money counts.
Best for: a bettor who needs to see the net price after commission.
Not ideal if: you would rather not do a second calculation for every price.
Value and account limits
Two kinds of text touch this, and they are set side by side here without a verdict. MadMarket's guide to value betting writes "Sustained success may lead to account restrictions with some bookmakers." Its FAQ adds "Some bookmakers may restrict or limit accounts that consistently take advantage of pricing inefficiencies." Both are MadMarket speaking about other books.
The brokers' own terms are a different thing. BetInAsia's terms reserve the right to refuse, restrict, cancel or limit a wager, and to adjust limits on individual accounts. The clauses of BetInAsia and MadMarket are set beside each broker's own words on the page on stakes and caps, and what applies on Orbit Exchange covers the exchange side. Nothing is repeated here.
Two things to read separately
MadMarket's lines above are about other books. The clauses on the linked pages are the brokers' own terms about their own accounts.
Questions about value, Kelly and exchanges
What is a value bet?
A value bet is a bet whose odds are higher than the true probability of the outcome, in the words of MadMarket's guide and of Sportmarket's FAQ. Your estimate of the chance is higher than the 1 ÷ odds the price implies: at 3.18 the price implies 31.45%, and an estimate of 34% leaves +8.12% per 1 staked (hypothetical).
Does a positive expected value mean the bet will win?
No. Expected value is an average over many bets, and one bet wins or loses. MadMarket's guide writes "Short-term variance remains, and losing streaks are still possible despite having an edge."
Which fraction of Kelly do people use?
The sources name different fractions. MadMarket's calculator offers Full, Half or Quarter Kelly and advises Half or Quarter "to reduce bankroll volatility and protect against over-estimating your edge". A Pinnacle article writes that 30 percent Kelly is a common figure. This site fixes none of them.
Can you work out value on an exchange?
Yes, with the commission taken off first. The price on an exchange is what other customers offer, and the commission comes off the net profit of the market. At an example rate of 3%, odds of 3.18 become 3.1146 and the expected value at an estimate of 34% falls from +8.12% to +5.90%. A bet that is never matched has no expected value at all.
Where to go next
The route from here is on the pro betting roadmap. When two prices add up to under 100%, the arbitrage page covers a different sum. What it costs to play at all is the subject of the running costs of a method, and the brokers' own calculators sit with the rest of the tools. To get to Orbit Exchange itself, read the routes in through a broker.