CLV Sample Size: How Many Bets Beat the Close?

A few hundred honestly logged bets with a positive average against Pinnacle's no-vig close is the start of evidence that you beat the closing line; results need thousands. CLV is measured on every bet, win or lose. Results are one noisy 0/1 per bet. That is why the sample for CLV is smaller than the sample for profit, and why a winning week still proves almost nothing.

This page is the sample-size question. For the formula and raw vs no-vig, start with closing line value. For how to log the bets, see the CLV tracker guide. For a one-week screen of a tool, see how to test a value betting tool.

Definition. CLV sample size is how many bets you need before the average (or the share) of closing line value is stable enough to trust. It is not a luck-adjusted profit target and it is not a promise that a positive CLV will show up as profit in the same window.

Why results need thousands and CLV needs hundreds

Each result is a coin with a small bias. Each CLV figure is a measurement of the price you got against the price the market closed at. The close can go with you or against you, but you get a number either way.

Take a bettor with a true +3% edge at odds of 2.00 (a 51.5% chance of winning each bet) and flat stakes. By plain binomial arithmetic, already published in betting discipline, the chance they are still below break-even after:

So after 500 bets, almost one bettor in four with a real +3% edge is still losing money. Judging the process by last month's P&L is how good rules get abandoned. CLV does not remove variance. It measures a different thing, on every trial. The same gap, as ROI versus CLV, is unpacked in sports betting ROI vs CLV.

How many bets for the share that beat the close?

Trackers often report two CLV numbers: the average size of the edge, and the share of bets that beat the close. They are not the same. The share is a proportion, so the sample size has a clean formula.

Assume, as a labelled working figure, that 55% of your bets have positive no-vig CLV (the rest are at or below the fair close). You want a two-sided 95% interval on that share to sit entirely above 50%, so 50% is no longer a plausible "coin flip".

n > (1.96² × p × (1 − p)) / (p − 0.50)²

These are original arithmetic from the formula above, not a study. Change the assumed share and the number moves a lot. A 52% process needs thousands of bets before you can tell it from 50%. A 60% process shows up in a couple of months of daily betting. If you do not yet know your share, you cannot pick a number from this table; you can only keep logging until the interval stops covering 50%.

A few dozen bets in a software trial can show a direction. They cannot exclude 50%. That is why how to test a value betting tool treats a week as a screen, not a verdict.

How many bets for the average CLV percentage?

The average needs the standard deviation of CLV per bet, which you only have after you log. Without it, nobody can convert "+2% average" into a sample size. Two honest rules:

• Do not invent a standard deviation. If your log does not yet have one, report the share (previous section) and wait.
• Once you have it: the standard error of the mean is s / √n. To get the standard error down to half a percentage point, you need n = (s / 0.5)² with s in percentage points. If your logged s is 8 points, that is 256 bets; if it is 12 points, 576. Those two s figures are illustrative, not measured on our users.

What a good CLV looks like, as opposed to how many bets you need, is already on the CLV guide: a positive average against Pinnacle's no-vig close, held over several hundred bets, with like-for-like markets. There is no official threshold.

What a small sample still lies about

Twenty bets. One steam move that holds can make the average look brilliant. One rebound can make it look broken. The CLV guide already flags this: twenty bets are not a sample.

A winning week with negative CLV. Variance. The process took worse prices than the close; the results came in. Do not scale the stake.

A losing month with positive CLV. Also variance. See the binomial table. Keep the rules; check that you are using the fair close and logging every bet, including the ones you skipped after the price moved.

Mixing markets. CLV on a Premier League 1X2 and CLV on an ITF games handicap are not the same quality of evidence, because the closes are not equally efficient. Track them separately as well as together.

Logging only the bets that felt good. Selection bias. If the log is incomplete, no sample size saves it. The bet tracking spreadsheet is the minimum; a tracker that captures the close automatically is better. Tools are compared in CLV tracking tools.

CLV sample size is not market-efficiency research

Joseph Buchdahl used 35,570 matches (106,710 1X2 prices, 22 European divisions, 28/07/2012–20/02/2017) to test whether Pinnacle's de-vigged prices predict actual returns at other books (football-data.co.uk, 24/02/2017, re-read 06/11/2026). That is a study of a reference. Your CLV sample is a study of your process against that reference. A few hundred of your bets can tell you whether your prices beat the close. They cannot re-run Buchdahl.

How Pinnacle Terminal handles this

Pinnacle Terminal's tracker records Pinnacle's closing price on every bet you log, from the app or from Telegram, and reports no-vig CLV with four devig methods plus a worst case (Power by default). You can export the taken price, the close and the result and recompute the share and the average yourself in the CLV calculator. It does not tell you that 380 bets is "enough" for your book; that depends on the share and the spread in your log. Live in-play football drops are shown and filtered, without a grade. The terminal does not take bets and is not affiliated with Pinnacle. Setup: tracker and CLV guide. Product: Pinnacle dropping odds. Plans: pricing.

FAQ

How many bets to know if you beat the closing line?

Enough for a 95% interval on the share of bets with positive no-vig CLV to sit above 50%. If that share is 55%, the arithmetic is about 380 bets. If it is 52%, about 2,400. Average CLV% still needs the standard deviation from your log.

Is 100 bets enough to judge CLV?

Usually not for a verdict. 100 results with a +3% edge at 2.00 still lose 34.4% of the time. 100 CLV figures can show a direction. They rarely exclude a coin-flip share.

Why does CLV need a smaller sample than profit?

Because every bet produces a CLV number, including losers. Profit is a 0/1 per bet with a small bias, so luck dominates for longer.

What if my CLV is positive and I am still losing?

Normal at a few hundred bets. Recheck the fair close, that every bet is logged, and that you are not mixing thin markets into the same average. Do not raise the stake to "catch up". Whether that close itself is a good probability estimate is is Pinnacle's closing line accurate.

Can I judge a 7-day trial by CLV?

You can screen it. A few dozen bets show whether the alerts you could place beat the close that week. They do not prove the tool, or you, beat the close as a process.

Sources

• Binomial table (+3% edge at 2.00; 100 / 500 / 1,000 / 2,000 bets): original arithmetic, first published in betting discipline (31/10/2026).
• Share-of-bets sample sizes (52% / 55% / 60%): original arithmetic from n > 1.96² p(1−p) / (p−0.5)², 06/11/2026.
• Joseph Buchdahl, "The Wisdom & Market Efficiency of Pinnacle.com compared to other Bookmakers", football-data.co.uk, 24 February 2017, re-read 06/11/2026 (35,570 matches).

Try it

To log every bet against Pinnacle's fair close automatically, start the Pinnacle Terminal 7-day free trial. Nothing is charged before day 8. Plans: pricing.

18+ · Bet responsibly. Only bet money you can afford to lose.