Closing line value: a diagnostic, not a verdict
Closing line value compares the price you took with a stated de-vigged closing price. It is a practitioner diagnostic, not realised ROI; the sources reviewed here do not establish long-run profit from it.
Learning objectives
- Compute closing line value against a raw and a de-vigged closing price, and explain why the de-vigged reference is preferred.
- Show how the de-vig method and the choice of 'close' change measured CLV, especially on long shots.
- Explain the practitioner reasoning behind CLV and state plainly what the evidence does not show.
- Quantify how far realised ROI can sit from CLV over a given number of bets.
What closing line value measures
The closing line is the last price before the event starts. Closing line value (CLV) asks: did you get a better price than the market's final view? The reference should be the fair (de-vigged) closing price, because the raw close still contains the margin.
CLV = o_taken / o_close,fair − 1If the fair closing probability were the true probability, CLV would equal expected value exactly, because p_close × o_taken − 1 = o_taken / o_close,fair − 1. That identity is the whole appeal of CLV — and its whole weakness, because it rests on the 'if'.
CLV is a property of when you bet as much as what you bet. A price taken early, before team news, carries more risk of moving either way; a price taken minutes before kick-off usually shows CLV close to zero. In the worked example, the same selection taken at 2.14 just before the close would have shown −1.27% against the proportional fair close. Compare CLV only between bets placed under the same timing rule.
Why practitioners use it — and what the evidence does not show
Timing and de-vig choices change the number
Which 'close'? The last snapshot before kick-off, the price at a market-leading book, or a multi-book consensus (lesson 7)? A thin book's close can be a poor reference. Record the timestamp of both the taken price and the closing snapshot, and use the same rule for every bet.
Which de-vig method? Proportional de-vigging spreads the margin evenly; Shin and power take more from long shots (Štrumbelj (2014) compares methods). On the hypothetical close 2.10 / 3.50 / 3.70, an away bet taken at 4.00 shows CLV of +8.11% against the raw close, +4.74% against proportional, +4.05% against Shin and +3.85% against power. The choice of reference can more than halve a long-shot CLV, so report which one you used.
| Reference | Home fair odds | CLV of home at 2.30 | CLV of away at 4.00 |
|---|---|---|---|
| Raw close | 2.100 | +9.52% | +8.11% |
| Proportional | 2.168 | +6.11% | +4.74% |
| Shin | 2.153 | +6.82% | +4.05% |
| Power | 2.148 | +7.05% | +3.85% |
CLV is not ROI: noise, selection and survivorship
Even if CLV equals true EV, realised ROI scatters widely around it. A bet at 2.30 with a 46.1% win chance has a profit standard deviation of about 1.15 units per unit staked, so 200 such bets leave a standard error of 8.11 percentage points on ROI. Positive CLV and a losing season are fully compatible.
- Selection: CLV is measured only on bets that were accepted. Prices you could not get filled, or stakes cut by limits, never enter the average.
- Survivorship: published CLV records come mostly from accounts that were not closed and bettors who chose to share. The losers rarely publish.
- Voided errors: bets struck at obviously wrong prices can show huge CLV and are then cancelled by the operator, so they flatter a record if kept.
- Reference drift: if your own bets move a thin market, the close partly reflects you, not independent information.
If you do track CLV, treat it as a sample statistic with its own uncertainty: report the average CLV, its standard deviation across bets, the number of bets and the share of bets with positive CLV. Track it separately by market, price band and time of placement, because a healthy average can hide one segment that consistently loses value. And remember that your own stakes can move a thin market, so a closing price that moved towards your bet partly measures you.
Worked example
Worked example · Measuring CLV on a Harbour City bet
On Friday you took a hypothetical 2.30 on Harbour City. The closing 1X2 at the reference book is 2.10 / 3.50 / 3.70. You plan 200 similar bets.
- Closing book sum
1/2.10 + 1/3.50 + 1/3.70103.22% (overround 3.22%) - Proportional fair close
Proportional de-vig46.135% / 27.681% / 26.185% → fair home odds 2.168 - Shin and power fair close
Shin de-vig (z ≈ 0.0161), Power de-vig (k ≈ 1.0307)Fair home odds 2.153 (Shin) and 2.148 (power) - CLV against each reference
2.30 / reference − 1Raw +9.52%; proportional +6.11%; Shin +6.82%; power +7.05% - Implied EV if the proportional close is true
0.46135 × 2.30 − 1+6.11% (identical to the proportional CLV, as the identity says) - Per-bet spread
mean and SD of one unit bet (p = 0.46135, o = 2.30)Mean +6.11%, standard deviation 1.147 units - ROI uncertainty over 200 bets
the standard error of ROI → 1.147 / √200SE 8.11 points; 95% band −9.78% to +22.00%; P(ROI < 0) ≈ 22.55% (normal approximation) - Over 1,000 bets
SE = 1.147 / √1000SE 3.63 points; P(ROI < 0) ≈ 4.60% - Bets to separate +6.11% from zero
bets needed to detect the edge (0.0611, 1.147)About 1,353 bets at 1.96 standard errors
The bet beat every reasonable closing reference, but the measured CLV ranges from +6.11% to +9.52% depending on choices you must report. Even if the close were exactly right, roughly one 200-bet run in four of such bets would still lose money. CLV tells you something about the price you took; the reviewed sources do not establish long-run profit from it.
Practice · Stress-test the closing probability
In the Fair Odds and EV lab, enter odds 2.30, probability 0.4614 (the proportional fair close) and a ±2-point range (0.4414 to 0.4814). Then widen the lower bound to 0.42 to represent a close that was itself too generous to Harbour City.
Lab 1 · Fair odds and expected value
Synthetic · deterministicPositive across your range. Even the low end of your range clears break-even. Under these assumptions the price looks favourable — but the range itself is an assumption, and margins, execution and correlated exposure can still remove the edge.
Show data table
| True probability (%) | Expected ROI |
|---|---|
| 34% | -22% |
| 35% | -20% |
| 36% | -17% |
| 37% | -15% |
| 38% | -13% |
| 39% | -10% |
| 40% | -8% |
| 41% | -6% |
| 42% | -3% |
| 43% | -1% |
| 44% | 1% |
| 45% | 3% |
| 46% | 6% |
| 47% | 8% |
| 48% | 10% |
| 49% | 13% |
| 50% | 15% |
| 51% | 17% |
| 52% | 20% |
| 53% | 22% |
| 54% | 24% |
| 55% | 26% |
| 56% | 29% |
| 57% | 31% |
| 58% | 33% |
| 59% | 36% |
Educational calculator. Uses only the numbers you enter or seeded synthetic data — no live prices, no staking, no real money.
Show what you should find
Break-even is 43.48%. At 0.4614 the EV is +6.12%, at 0.4414 it is +1.52% and at 0.4814 it is +10.72%: positive across this range. Moving the lower bound to 0.42 gives −3.40% at the low end, so the verdict becomes 'depends on the estimate'. Positive CLV is only as good as the assumption that the closing price is close to the truth.
Common mistakes
- ✗ Measuring CLV against the raw closing odds.The raw close includes the margin. Use a de-vigged close; in the example raw CLV (+9.52%) overstates proportional CLV (+6.11%).
- ✗ Saying CLV is proven to predict profit.The reviewed library does not establish that claim. The reasoning depends on the close being efficient.
- ✗ Expecting realised ROI to match CLV over a season.With 200 bets at 2.30 the ROI standard error is about 8 points; a loss is quite possible despite positive CLV.
- ✗ Reporting CLV without stating the close and the de-vig method.Timing and method change the number, most on long shots (+4.74% vs +3.85% on the away example). Fix and report both.
When it fails
Supporting research
- Simon (2024): Inefficient Forecasts at the Sportsbook: An Analysis of Real-Time Betting Line MovementManagement Science · Peer-reviewed empiricalEvidence (MLB, four US books) that line movement can overreact and that forecast quality does not always improve toward game time, so the close is not guaranteed to be the best forecast.
- Sauer (1998): The Economics of Wagering MarketsJournal of Economic Literature · Literature reviewThe efficiency view of wagering prices on which the practitioner case for CLV rests.
- Štrumbelj (2014): On determining probability forecasts from betting oddsInternational Journal of Forecasting · Peer-reviewed empiricalDifferent de-vig methods give different fair probabilities, which changes measured CLV.
Check your understanding
Practice only: saved on this device, no XP and no account record.