Prediction market closing line value is a narrow market-relative diagnostic: it compares an entry price with a consistently defined reference close. A positive number means the entry finished ahead of that reference for the chosen side. A negative number means the entry finished behind it. That comparison is useful, but it is not a profit proof, a realized return, or an expectation estimate by itself.
ZenHodl's descriptive CLV scorecard provides a practical example of why the close definition matters. We corrected an outcome-contaminated in-play comparison in 2026; the audit below explains the mistake.
The definition and sign convention
For a prediction-market BUY, CLV in cents is:
reference close − entry price
For a SELL, the sign reverses:
entry price − reference close
A buyer is ahead when the reference close is above the entry. A seller is ahead when the reference close is below the entry. Positive CLV therefore means the entry beat the chosen reference after accounting for the side. Polymarket CLV follows the same buy/sell convention when prices are quoted in cents.
The reference close must be defined before looking at outcomes. A final pregame close is one definition. A near-terminal in-play price is another and can encode the eventual result. A resolution settlement is not the same thing as the reference close.
A hypothetical worked example
Use the following arithmetic as an illustration, not as a recorded trade. Suppose a market is bought at 35¢, and the defined reference close is 41¢.
- BUY:
41 − 35 = +6¢ - SELL at the same 35¢ entry and 41¢ reference:
35 − 41 = −6¢
For a $1-settling contract, that six-cent difference corresponds to six percentage points in its quoted probability. It is not a 6% return. Even the entry-relative quote change, (41 / 35 − 1) × 100 = 17.14%, describes a price move rather than realized P&L.
For a sportsbook decimal-odds comparison, an entry at 2.10 and a reference close at 2.00 gives (2.10 / 2.00 − 1) × 10000 = 500 basis points. That is a 5% relative odds comparison, not a 5 percentage point change in win probability.
Calculator arithmetic in code
The checked Python block below keeps the sign convention and domain checks strict. Prices must be strictly between 0 and 100 cents. Decimal odds must exceed 1. The expected outputs are 6, -6, and 500.00.
from decimal import Decimal
def clv_cents(entry, reference, side="buy"):
entry, reference = Decimal(str(entry)), Decimal(str(reference))
if not (0 < entry < 100 and 0 < reference < 100):
raise ValueError("Enter prices strictly between 0 and 100 cents")
if side == "buy":
return reference - entry
if side == "sell":
return entry - reference
raise ValueError("Side must be buy or sell")
def sportsbook_clv_bps(entry_decimal, closing_decimal):
entry, close = Decimal(str(entry_decimal)), Decimal(str(closing_decimal))
if entry <= 1 or close <= 1:
raise ValueError("Decimal odds must exceed 1")
return (entry / close - 1) * 10000
print("buy CLV (cents):", clv_cents(35, 41))
print("sell CLV (cents):", clv_cents(35, 41, "sell"))
print("sportsbook CLV (bps):", sportsbook_clv_bps("2.10", "2.00"))
You can also compare entries and reference prices with the free closing line value calculator.
Pregame and in-play references are different
For pregame CLV, choose a consistent last usable price before the event starts. Record its source and clock, and decide how old a quote may be before it becomes unusable.
For in-play work, a fixed-horizon markout is a different declared reference: compare the fill price with a price at a specified elapsed horizon after the fill. Set the horizon in advance and apply a separate quote-age limit. Record score-event timing as well; a price after a major event can reflect information that was unavailable at entry.
Keep attempted orders and non-fills separate from filled observations. Report missing references, fees and the population included in each measure. Separate venues and units. For uncertainty estimates, account for dependence within games rather than treating every contract as an independent observation.
ZenHodl's dated negative evidence
The public correction record documents the following historical audit. It is not a new October 7 measurement.
In a May 8 snapshot, ZenHodl reported win rates of 89.9% versus 11.2% after splitting trades by the sign of CLV. We retracted the forecasting interpretation on July 21, 2026: the near-terminal in-play close already encoded much of the eventual outcome. Grouping entries by whether they beat that close created an impressive-looking split without establishing forecasting skill.
The test that matters replaces entry rules while keeping captured closes and outcomes fixed. The published audit used flat 50c entries, shuffled entries and coin-flip decisions. All reproduced large gaps. In the flat-50c threshold sweep, the reported gap rose from 88.8 to 95.3 percentage points as the CLV threshold tightened. The retraction page preserves these placebo results.
An August 6 amendment corrected another part of the argument: the inverse-side gap was an arithmetic identity, not an independently observed opposite-side book or an empirical no-skill control. The flat-price, shuffled-entry and coin-flip placebos are the falsifiable checks.
This finding applies to our outcome-contaminated in-play close and the inference drawn from it. It does not invalidate a consistently defined pregame comparison. Either reference still needs independent evaluation before being used to support a claim of forecasting skill or profitability.
To record your own entries, use the bet journal and grader or Kalshi grader. Keep the chosen close definition with the observation so later comparisons use the same rule.
Method note, October 7, 2026: DeepSeek assisted drafting. We reran the Python arithmetic and checked the audit summary against the dated public correction. The price examples are hypothetical; the placebo findings are previously published evidence.