← Learn

12 min read · updated September 26, 2026

Closing-line value: how to know if you're good before the results say so

Wins and losses take hundreds of bets to tell skill from luck. The price you got, compared with where the market closed, starts telling you something after a few dozen. What closing-line value is, how to work it out, and what the site's bet tracker does with it.

The short version

A bet's result is mostly luck. Even a genuinely good forecaster can lose for a month, and a bad one can win for a season, so the scoreboard needs hundreds of bets before it says anything reliable. There is a faster signal. Just before kickoff, the betting market settles on its final price, the closing line, and that price is the best public forecast of the game there is. If the prices you take are consistently better than where the market closes, you are seeing something the market later agreed with. That gap is closing-line value, or CLV. It doesn't prove you're good, but it's the best early sign there is, and it starts to mean something after a few dozen bets instead of a few hundred.

This is about how prices and forecasts work, not advice to bet. If you do bet, bet only where it's legal, only if you're of age, and only within your means.

Why results are so slow

Start with the simplest bet there is: a game both sides agree is a coin flip, priced at −110 on each side. That price charges 110 / (110 + 100) = 52.4%, so to break even you have to win 52.4% of the time. The extra 2.4 points on each side is the cut, the book's margin, also called the vig, the juice or the hold. (The odds article covers where it comes from.)

Now imagine a bettor who is genuinely good: on these bets, he truly wins 54% of the time. Per dollar at −110 (decimal odds 1.909), his expected return is 0.54 × 1.909 − 1 = about +3¢. A 50% bettor at the same price expects 0.50 × 1.909 − 1 = about −4.5¢. The gap between good and ordinary is a few cents a dollar.

How long before his record shows it? Each bet is a win or a loss, and a win rate measured over n bets wanders around its true value by about this much (one standard error):

standard error of a win rate = √( p × (1 − p) / n )

With p near 50%, the top of that is √0.25 = 0.5, so the standard error is 50 points divided by √n:

The ranges are for our 54% bettor. After 100 bets, anything from 44% to 64% is an ordinary outcome for him, and that range swallows both the break-even line (52.4%) and a coin flipper (50%). After 400 bets it still includes both. To tell a 54% bettor apart from a 52.4% one, a gap of 1.6 points, the standard error has to shrink to 1.6 points before the gap is even one standard error wide: 0.25 / 0.016² = about 980 bets. Two standard errors, the usual bar for "probably not luck", takes four times that, near 3,900.

That's the uncomfortable truth about results. A small edge takes hundreds of bets just to glimpse and thousands to confirm. Most bettors lose over time and the cut is the main reason, but in any given month plenty of them are up, and they can't tell from the record which kind they are.

What the closing line is

Prices move between the moment a line opens and kickoff. Injury news arrives, weather forecasts firm up, and money comes in from people with every kind of opinion. By kickoff, the price has absorbed all of it. That last price is the closing line.

On this site, the close is each market's last reading before kickoff (the sportsbooks, Kalshi and Polymarket), averaged together, with the cut removed. What's left is a fair probability: the market's final, best estimate of each side's chance.

Why trust it? Because it's very hard to beat. Scored the way this game scores picks, over 6,918 games from 1999 to 2025, the closing line averages 997 points a season. Elo trails it by more than 100 a season. The System, which starts from the market and adds a few rules chosen on those same 27 seasons, is ahead by 14 a season with a standard error of 10, ahead in 16 of the 27 seasons: even, inside the noise. If a forecast built on the close, and tuned with hindsight, can't clearly beat it, the close is a very good yardstick. The market article goes into why.

What closing-line value is

The idea is simple: compare what you paid with what the close says the bet was worth.

Every price charges a probability. DET at +150 charges 100 / (150 + 100) = 40.0%. Taking that bet is saying, in effect, "I think DET's chance is better than 40%." The close then gives the market's final answer. CLV is the difference:

CLV (points) = closing fair probability for your side − probability your price charged
CLV (return) = closing fair probability ÷ probability your price charged − 1

The points version is easy to read. The return version answers a more useful question: if the close is the true chance, how much would this bet make per dollar, on average?

Worked examples

An underdog that moved your way

You take DET at +150, which charges 40%. By kickoff, the market has moved toward DET, and the close says 44% fair.

• Points: 44 − 40 = +4 points.
• Return: 44 ÷ 40 − 1 = +10%.

Check the return another way. +150 is a decimal payout of 2.5, and if DET's true chance is 44%, the expected return is 0.44 × 2.5 − 1 = 0.10, or 10¢ a dollar. Same answer. That isn't a coincidence: the probability a price charges is 1 divided by its decimal payout, so "close ÷ charged − 1" and "chance × payout − 1" are the same sum. CLV as a return is the bet's expected value, measured against the close.

DET might lose. It will lose more often than it wins. The bet was still a good price, by the market's own final judgment.

A favorite that never moved

You take the home team at −135, which charges 135 / (135 + 100) = 57.4%. The other side is +115, which charges 46.5%. Together they add to 103.9%; the extra 3.9% is the cut. Remove it and the fair numbers are 57.4 / 103.9 = 55.3% for the home team and 44.7% for the away team.

Suppose nothing happens all week and the close lands exactly there, 55.3% fair.

• Points: 55.3 − 57.4 = −2.1 points.
• Return: 55.3 ÷ 57.4 − 1 = about −3.7%.

The line didn't move against you, and you still have negative CLV. That's the cut. Whenever you pay a book's price, you start behind the fair number, so just matching the market isn't enough. To be ahead you have to beat the close by more than the margin you paid.

The classic coin flip

At −110 on a game that closes 50/50: 50 − 52.4 = −2.4 points, and 50 ÷ 52.38 − 1 = about −4.5%, the same 4.5¢ a dollar as the break-even arithmetic above. If you always bet right at kickoff, your price is the close, plus the cut, and your CLV will sit at about minus the cut forever. The only way to positive CLV is to get a price before the market gets there, or a better price than the market average.

A parlay

A parlay multiplies. Its CLV is the product of each leg's closing fair probability, divided by what the payout charged, minus 1.

Take two legs at −170 (decimal 1.588) and −110 (decimal 1.909). The parlay pays 1.588 × 1.909 = 3.032, which charges 1 / 3.032 = 33.0%. If the legs close at 60% and 50% fair, in separate games, the close says the parlay's chance is 0.60 × 0.50 = 30%.

• Points: 30 − 33.0 = −3.0 points.
• Return: 0.30 × 3.032 − 1 = about −9%.

A single at −110 on a coin flip starts about 4.5% behind; this two-leg parlay starts about 9% behind, because the cut compounds with every leg. (The multiplication only works for independent games; two legs from the same game move together, and the product is wrong.) The parlay article has more.

The four, side by side

Why CLV speaks sooner

A result is all or nothing. A 50% bet either wins or loses, so a single result lands 50 points away from its expected value every time, and the noise only averages out slowly, as the table above shows.

CLV doesn't wait for the game. It measures the price you got against a forecast, and the difference is a few points, not a whole win or loss. Much less noise per bet means far fewer bets before a pattern means something. As a general rule, a few dozen bets of CLV tell you more than a few dozen results, and consistently beating the close is the best early sign of an edge. Results still settle the question in the end; CLV tells you which way the evidence is leaning while you wait.

It also points at the right thing. A bettor with positive CLV and a losing record has probably been unlucky. A bettor with negative CLV and a winning record has probably been lucky, and the record is likely to drift back. That second case is the more common and the more dangerous one, because a winning month feels like proof.

Honest limits

• CLV trusts the close. It measures you against the market's last price. If the market was wrong about a game, CLV can't see it. Over 27 seasons it has been very hard to beat, but hard isn't impossible, and on a game where the close was wrong, CLV will mark a right call as a mistake.
• The close here is an average. Book prices are read every six hours, so a book's close can be a few hours old; exchanges are read every 20 minutes. A late injury can move the true close after the last book reading.
• Parlays need independence. The parlay CLV multiplies each leg's close, which is only right for legs from different games.
• It isn't a guarantee. Positive CLV means you've been getting good prices by the market's own judgment. It doesn't promise that you will keep doing so, or that any bet will win.

Results against expectations

The other half of the picture is comparing what happened with what the close expected. Say you've placed 50 singles, and their closing fair probabilities average 45%. The close expected 50 × 0.45 = 22.5 wins. If you won 26, you're 3.5 wins ahead of the close's expectation. Is that skill? With 50 bets, the standard error on a count of wins is √(50 × 0.45 × 0.55) = about 3.5 wins, so you're one standard error ahead, well within luck. Your CLV on those same bets tells you whether the prices were good; the results tell you how the dice fell.

Calibration asks the same question in bins. Group your bets by their closing number: the ones that closed around 40% should have won about 40% of the time, the ones around 60% about 60%. If they did, the close was right about your bets and your results are mostly luck either way. If your 40% closes have been winning far more than 40% over many bets, that's worth noticing, and it's also exactly where results take the longest to trust.

Where to see it on the site

• Bets: the bet tracker. Log a single or a parlay with the price you got. It settles each bet from the final score and shows your totals, your ROI (profit divided by what you staked), your CLV, your results against what the close expected, breakdowns by singles and parlays, favorites and underdogs, and books and exchanges, and a calibration table by closing number. You can publish your record at /bets/your-username; stake amounts aren't shown.
• Price check: before a bet, type a price and see what it charges, the fair number, and whether it's under or over. It's the same comparison as CLV, made before kickoff instead of after.
• Markets: every game's home win percent from the books, Kalshi and Polymarket, the biggest gaps between them, and line moves, the movement CLV measures.
• The System and the calibration page: how the forecast you're trying to beat is built, and how well Elo's numbers have matched outcomes.
• For the background: how betting odds work, why the market is so hard to beat, and scoring, confidence and calibration.

A few terms

• Closing line (the close): the last price before kickoff. Here, each market's last reading, averaged, with the cut removed.
• Fair probability: a price's implied chance with the cut taken out.
• The cut: the margin built into a price, which makes both sides add to more than 100%. Also called vig, juice or hold.
• CLV: closing-line value. The closing fair probability for your side minus what your price charged, in points, or as a return, the close divided by the charge, minus 1.
• Standard error: how far a measured number usually wanders from its true value by chance. For a win rate, √(p × (1 − p) / n).
• ROI: return on investment. Profit divided by total staked.
• Calibration: whether things forecast at 70% happen about 70% of the time.