Polymarket Implied Probability, Explained

• A Yes token priced at 0.p means p percent probability in the simplest reading.
• The implied probability of No is 1 - Yes price; the two tokens must sum near 1.00 on a two-outcome market.
• Mid price is the naive probability; best ask is the probability you can actually buy, and best bid the one you can sell.

Recorded books show mids that are not executable probabilities. A mid of 0.4995 resting on an empty book, or a 0.500 printed over a 99.6-cent-wide chasm, quotes a probability nobody can trade at.

For tradeable probability, read the book: the ask is the cost of Yes, the bid is the proceeds of selling. The spread between them is the friction on the estimate.

Near resolution the implied probability collapses to the answer: the winning side holds 0.99 or better in 98.6% of measured observations, and the losing side drains toward 0.01. The probability quoted by the book stops being an estimate and becomes the outcome.

Studies of these moves need the recorded ladder — a 1-minute price series shows the collapse as a step; the 250ms book shows the depth that preceded it.

• Take an Up/Down frame and compute the implied probability from the executable floor (floor = the market's probability of the up side) and from the mid; the two disagree on one-sided books, and the gap is information.
• Recompute the same implied probabilities five seconds later after a reprice; the one-second 99.6-cent ladder pass shows how fast anchor probability moves.
• Compare the implied probability to the actual settlement frequency across a sample of resolved contracts; calibration is the real test of an implied number.
• Record the bound and reference time alongside every implied reading; an implied probability without its reference is a snapshot without coordinates.

Implied probability on Polymarket is a market price, not a forecast: it is the venue's answer to "how much is each side worth," and on a one-sided, thin ladder it can be a one-tick statement about a fast-moving reference.

The measured calibration corpus — 98.6% of frames at 0.99/0.995, 90%+ one-sided — means implied readings bunch at the edges, and the interesting inferences sit in the compressed 0.50–0.60 band where the ladder changes fastest.

Calibration is the honest test of any implied number on this venue: take a sample of resolved contracts, compare the implied floor at a chosen second to the true settlement share, and publish the calibration curve with its sample size rather than a single point.

A practical sanity check for any implied-to-value conversion on this venue: convert the floor to a probability, convert the odds quote to a price, and confirm they agree within one tick. The two readings should describe the same ladder, and when they diverge the mid is usually the culprit rather than the research, which is why every archived frame stores the floor and the reference second together.

Implied probability is the venue's central estimate in price form, valid only with its reference time and bound.

Check calibration, not just level: a series of resolved contracts is the honest test of any implied claim.

The tied-to-the-ladder reading — floor-based, one-sided-aware — is the one that survives a reprice audit.

Implied probability is the read-out of the Up/Down ladder: this page converts the measured price structure into the numbers people quote.

Pair it with the odds and mid-price explainers; the three pages cover price, probability, and executability of the same frame.

Calibration, not level, is what this page adds to the catalog; the resolved corpus is the reference for its tests.

An implied probability is a price with a reference time and a bound; without them it is a number without coordinates.

Check calibration on resolved contracts before believing any implied edge; the corpus is the reference.

• Order book data
• Price history
• Crypto data
• Provider matrix