Strategies·November 17, 2026·7 min read

Closing Line Value on Polymarket: The Scoreboard That Beats Profit

How to measure CLV on Polymarket, why a profitable record can hide a losing process, and the sample sizes needed to keep or kill a strategy.

closing line valueCLVstrategy evaluationpolymarketsample size
Calculator and math — illustrative image
Photo: Michal Jarmoluk / StockSnap · CC0Source

Closing line value is the difference between the price you got and the price the market settled on just before the event resolved. If you bought Yes at 42 cents and it closed at 50, your CLV is 8 cents. If it closed at 38, your CLV is minus 4. The event outcome is irrelevant to the measurement, which is why it is useful.

Profit tells you what happened once. CLV tells you whether you were on the right side of the market's final information. In a binary market, one trade can be right for the wrong reason and one can be wrong for the right reason. Over a few hundred trades, CLV separates process from luck.

What CLV measures

The closing price is the market's last collective estimate before the outcome is known. For sports, that is usually kickoff. For politics, poll close or the scheduled resolution time. For a data release, the price just before the release. The clock must be fixed in advance.

CLV per share is the closing price minus your entry for a long, or your entry minus the close for a short. A 42 cent entry that closes at 50 is plus 8 points, or 19 percent of the stake.

The close is not the truth. It is another estimate, made with more information and usually by a deeper market. Positive CLV means you were earlier or better, not right. Negative CLV is evidence against your process, not proof.

The arithmetic, with fees

Take a politics contract bought at 42 cents. The taker fee is 0.04 times 0.42 times 0.58, or 0.97 cents, so the all-in cost is 42.97 cents. If it closes at 50, gross CLV is 8 cents and net CLV is 7.03 cents, or 16.7 percent of the stake. If it closes at 38, gross CLV is minus 4 cents and net is minus 4.97 cents, or minus 11.8 percent.

A maker entry changes the arithmetic. The fee is zero, but the fill is not free: you get it when someone sells to you, which is more likely when the market is about to move against you. Compare maker and taker CLV separately.

Why profit alone is noisy

Suppose you make 100 trades at 50 cents with a $100 stake and win 52. Profit is 52 times $100 minus 48 times $100, or $400, a 4 percent return on $10,000 of turnover.

The standard deviation of the number of wins is the square root of 100 times 0.52 times 0.48, or about 5 wins, or $500. A 52 percent win rate is less than one standard deviation from a coin flip, so the $400 is indistinguishable from noise. You need hundreds of trades to tell a 52 percent edge from 50.

CLV is more informative because it is a continuous price difference, not a binary outcome. If per-trade CLV has a mean of 1.5 cents and a standard deviation of 8 cents, the standard error after 100 trades is 0.8 cents, a t-statistic of 1.9. After 200 it is 0.57 cents and t is 2.6. Profit can need thousands of observations; CLV usually needs hundreds.

A worked sample

Imagine 120 trades of 100 shares, average entry 44 cents, average close 46.5 cents. Gross CLV is 2.5 cents per share. The politics fee is 0.04 times 0.44 times 0.56, or 0.99 cents, so net CLV is 1.51 cents, or 3.4 percent per trade.

Total net CLV is 120 times 100 times 0.0151, or $181.20, on $5,280 of deployed capital, though not all of it is committed at once. Suppose realized profit is minus $40. The outcomes were unkind: several 30 cent positions lost and the winners were small. CLV says the entries beat the closes; profit says the sample was unlucky. Both can be true. Killing on the $40 loss may kill a positive process; scaling on CLV alone may ignore an exit or sizing problem.

Look at the distribution too. If the mean is positive but the median is negative, a few outliers drive the result. If the beat-close rate is 58 percent and the mean is positive, the evidence is more consistent. If one trade closed 30 points above entry, you have one observation.

Sample size and confidence

For a per-trade CLV standard deviation of 8 to 10 cents, a few hundred trades are usually needed to get the standard error below half a cent. If your edge is 1 cent, that is the difference between a t-statistic of 2 and one of 1. If the edge is 3 cents, 100 trades may be enough.

Segment before pooling. Politics and sports have different fees, clocks, and variance. A positive pooled average can hide a negative category. Require the edge in two independent segments, or in one segment across two periods, before scaling. Beware multiple comparisons: testing ten variations and keeping the best CLV is fitting noise.

Keep or kill

Keep if mean net CLV is positive, the t-statistic is at least 2 over 200 or more trades, and the sign is stable across two segments or periods.

Investigate if profit is positive but net CLV is negative. That usually means lucky outcomes, or a risk the closing price does not capture, such as resolution risk or a concentrated tail.

Kill if mean net CLV is below minus 1 cent over 150 or more trades, if CLV is positive only before fees, or if the edge disappears when you exclude the largest winner.

Do not kill on a drawdown alone; a positive-CLV strategy can have a 20 percent drawdown. Do not keep one on a hot streak alone; a negative-CLV strategy can have a great month.

Measurement traps

The first trap is the clock. If you measure the close at a different time for each trade, you measure different things. Fix it by category before the sample starts.

The second is the midpoint. Polymarket shows the midpoint unless the spread is wider than $0.10, when it shows the last trade. A midpoint close can be unavailable in a thin book, and a last-trade close can be stale. Use the executable side, or use the midpoint consistently and disclose it.

The third is your own impact. If your order moved the price, the post-trade midpoint is not independent. Use the pre-trade midpoint, or exclude trades where your size was large relative to the book.

The fourth is resolution. A contract that closes at 50 and resolves 50/50 has 8 cents of CLV and a zero return. Keep resolution risk in a separate column.

How to execute

  1. Define the closing clock for each category and never change it mid-sample.
  1. Log entry time, side, size, fee, and executable entry price.
  1. Log the closing price at the fixed clock, using the same side or midpoint convention.
  1. Compute net CLV per share and as a percent of stake, then aggregate by category and month.
  1. Track the beat-close rate and the median alongside the mean.
  1. Review at 100, 200, and 400 trades, and do not change the strategy between reviews.

Risks

The biggest risk is treating CLV as a guarantee. Positive CLV can coexist with a losing year, and negative CLV with a winning one. CLV is a leading indicator, not a settlement.

The second is a distorted close. In a thin market, the last trade before resolution may be one small order. In a wide market, the midpoint may be untradeable. A CLV number from bad closes is worse than no number because it looks precise.

Checklist

  • Is the closing clock fixed in advance for every category?
  • Is the closing price executable, or is it a midpoint I have disclosed?
  • Have I netted out fees and used the same side for entry and close?
  • Do I have at least 200 trades, or a clear reason to accept fewer?
  • Is mean CLV positive after fees, with a t-statistic above 2?
  • Is the edge stable across two segments or two time periods?
  • Am I keeping the strategy for CLV reasons, not because the last month was profitable?

Closing line value is not a perfect measure. It is the best cheap measure of whether your entries beat the market's final information. Profit is the outcome; CLV is the process. A trader who tracks both, and knows which to trust at which sample size, is much more likely to keep the right strategy and kill the wrong one.

Sources

  1. 1.Beating the bookies with their own numbers— arXiv
  2. 2.Explaining the Favorite-Longshot Bias: Is it Risk-Love or Misperceptions?— National Bureau of Economic Research
  3. 3.Prices & Orderbook— Polymarket Documentation
  4. 4.Fees— Polymarket Documentation
  5. 5.Resolution— Polymarket Documentation

More from the journal

All articles