Value betting explained: fair odds, edge and closing line value

What value means

Every decimal price implies a chance: one divided by the odds. A price of 2.00 implies 50%. If the true chance of the event is 55%, then over many such bets a pound staked at 2.00 returns £1.10 on average — the price is value. If the true chance is 45%, the same price is a losing one, however often it happens to win this week. Value is a statement about the price, not about the result.

That is also the whole difficulty. Nobody knows the true chance. A value bettor is anyone with a better estimate of it than the price implies, and the honest question is always: better according to what?

Where a fair price comes from

There are two answers. One is a model — a projection built from the data, of the kind Per90 uses to price player props and describes on its methodology page. The other is the sharp market: a bookmaker that runs a low margin, accepts large stakes from winning customers, and moves its price on the money it takes, so that its price is the best public estimate of the truth that exists. In football that is Pinnacle. Its price is not the truth, but it is what the rest of the market is measured against, and a UK bookmaker’s price that stands above it is making a claim the sharpest money in the world disagrees with.

A sharp price still carries a margin — the overround. Removing it is called de-vigging: the implied chances of every outcome are scaled back until they add up to 100%. What remains is the fair price.

The edge is the expected return per unit staked if the fair price is right. Every value bet carries that condition, and the rest of this guide is about how to find out whether it held.

What an edge is not

An edge of 5% is not a 5% chance of winning, and it is not 5% profit in hand. In the example the bet still loses 50% of the time. Most value bets lose individually; the claim is that across many of them, at prices that were genuinely too big, the wins outweigh the losses. A single result says nothing about whether the price was right, and a month of results says very little.

Nor is an edge a fixed thing. It exists at a moment, between two prices, and either can move. A pick built on a reference price that has since moved is a claim about a market that no longer exists, which is why every pick on this site carries the age of both prices, not just one.

The test: closing line value

If results cannot tell you whether a price was right, what can? The market’s own final verdict. The closing line is the sharp price just before kick-off, when the most information and the most money have moved it. If the prices you took consistently beat the close — you backed at 2.10 and it closed at 1.95 — you were finding real value, whether or not the bets won. If you took 2.10 and it closed at 2.20, the market disagreed with you, and it is usually right.

Closing line value settles in hours rather than seasons and is far less noisy than results, which is why the Per90 public record shows it beside every pick’s detected edge and its settled result. A big detected edge with flat or negative CLV is finding noise. The same test removed a whole reference from the board: Betfair Exchange had been the fair price for corners, both teams to score, double chance, draw no bet and shots on target, and over 634 picks those closed at −4.25% against the line while Pinnacle-referenced picks closed at +0.11%. The exchange was withdrawn on 1 September 2026 and its picks were removed from the current-rules record. That is what the test is for.

Staking

The Kelly criterion gives the stake that grows a bankroll fastest if your edge is exactly right: the edge divided by the price minus one, as a fraction of the bankroll. Nobody’s edge is exactly right, and full Kelly on an overestimated edge is how bankrolls disappear. The usual answer is a fraction of it. Per90 shows quarter Kelly, capped at 2.5% of the bankroll you enter, and the bankroll never leaves your browser. Level stakes — one unit on everything — are the other honest choice, and the public record is graded that way so it can be read without a staking argument.

How the Per90 value board applies this

• The reference is Pinnacle, de-vigged, on match result, total goals, Asian handicap and goal lines. Nothing else is published. Whole and quarter lines are priced only when the adjacent half-lines pin down the outcome, and a push or half-stake return is counted in the edge.
• Four UK books are scanned — Bet365, Sky Bet, Paddy Power and William Hill — every 5 minutes. A pick is one of them offering at least 5% expected return per unit staked at the reference chances, or 3% on match result, where the market is too tight to reach the higher bar.
• At most two picks per match since 5 September 2026: one on total goals, one on the match result or handicap. The first qualifying selection keeps its slot for the life of the fixture; a later, better-looking line cannot replace it.
• Every pick is graded against the closing price and the final score, in units and CLV, and kept: losing runs, withdrawn calls and voids stay on the record under the same rules as the wins.

Practical rules

1. Singles. Combining value bets multiplies the margin you pay and the variance you carry.
2. Check the price is still there before staking, and note the age of the reference price.
3. Record every bet with the price taken; add the closing price afterwards.
4. Judge yourself on CLV first, units second, and neither on fewer than a few hundred bets.
5. Expect restrictions. Bookmakers limit accounts that consistently beat the close, and nothing on this site changes that.

What this does not do

The reference is the market’s best public estimate of the truth, not the truth. An edge can lose, most do individually, and a record this young swings far more in units than in CLV. Per90 publishes the comparison and the record so the claim can be judged; it does not promise the claim holds, and no outcome is guaranteed. Research on published prices, not betting advice.

Open the value board → for what stands above fair now, or the record for how every pick has closed.