How to see Polymarket historical odds for free

Four routes to past odds, and what each one costs you

Reading where a Polymarket price has been is a different job from reading where it is now, and the tools split along that line. If the question is casual, Polymarket's own market page already draws the line and needs nothing from you. If the question is precise, the public API returns the underlying points. If the question reaches back months, neither of those is enough on its own, and that is the part most write-ups skip.

One quirk of searching for this is worth knowing. The phrase Polymarket historical data returns a large number of cryptocurrency exchange pages offering price history for tokens that happen to carry the Polymarket name. Those are unrelated to the prediction market and to the odds described here. The prediction market publishes its own data through the API below.

What the free endpoint actually returned when we tested it

The endpoint is https://clob.polymarket.com/prices-history. Its documentation lists a required market parameter, which is the CLOB token id of a single outcome rather than the market slug, plus optional startTs, endTs, interval and fidelity values, and states that no authentication is needed. Because one token id means one outcome, a YES series and a NO series are two separate requests, and on a standard two-outcome market the two lines are mirror images that sum to a dollar.

Documentation describes the shape of an API but rarely its edges, so on August 1, 2026 we sent the requests below against the YES token of a long-running open market and recorded what came back. The point counts and spacing are measured, not quoted.

Two behaviours in that table are worth stating plainly, because both send people in circles. First, the intervals are not free-form: each range enforces a minimum spacing, so the one-month range refuses anything finer than ten minutes and the one-week range refuses anything finer than five. A bare interval=1m looks broken and is in fact a missing fidelity value. Second, the floor wins over the request. Passing fidelity=1 alongside interval=max still returned points ten minutes apart.

Why all does not mean all time

The headline finding is the one that changes what anyone can build. Both interval=all and interval=max returned about 31 days, and the earliest point in every response was dated July 1, 2026. We repeated it on three markets, including two that had been open since 2025, and each returned roughly 4,400 points beginning on that same July date. The window looks rolling rather than complete: what comes back is the last month or so, regardless of how long the market itself has existed. Explicit timestamps do not get around it either, since a startTs and endTs pair 20 days apart was refused as too long while 14 days was accepted.

The practical consequence is that a chart of a Polymarket price across an entire election cycle cannot be assembled from the public endpoint after the fact. It has to have been recorded while it happened. That is why the independent archives exist and why some of them charge: their product is not access to the data, which is free, but the fact that somebody was collecting it a year ago. Anyone who expects to want a long series later is better served starting a small scheduled job now than hoping to backfill it, and API behaviour like this is exactly the kind of thing that shifts without announcement, so it is worth re-checking rather than trusting a dated write-up including this one. The wider developer picture sits in does Polymarket have an API.

Reading a price series without over-reading it

A past odds line is a record of what the market charged for an outcome, not a record of what was true. A contract that traded at 30 cents for a month and then resolved NO was not necessarily wrong, in the same way that a forecast of rain that stayed dry was not necessarily wrong. The only honest test of calibration runs across many markets at once, which is the subject of how accurate is Polymarket, and the reading of a single price as a probability is covered in how to read implied probability.

Thin markets add a second trap. A price point exists because a trade happened or a quote sat there, and in a market with little resting depth a single small order can move the printed line several cents and leave a dramatic-looking spike that represents very little money. Reading a series alongside volume and liquidity is what keeps that in proportion, and the difference between those two numbers is unpacked in volume vs liquidity on Polymarket. The reasons a line moves at all are gathered in why do Polymarket prices move.

What Cent Signals shows, and what it does not

It is worth being direct about our own limits here, since this page is about finding history and this site does not publish one. Cent Signals builds from a dated snapshot of the public Polymarket data: current prices, volume, liquidity and the visible positions of large wallets. A market page such as will the US invade Iran before 2027 shows where pricing stood as of that snapshot, not the path it took to get there. Every figure on the site carries the snapshot date for that reason, and the cadence and its limits are set out on the methodology page.

Where a snapshot does earn its keep is cross-sectional reading, meaning the comparison of many markets at one moment rather than one market across time. The markets index and the traders index are built for exactly that. Wallet-level history is the one place the public record is genuinely deep, because settled fills live permanently on-chain rather than in a rolling window, which is what makes large-wallet tracking possible at all. That route is described in how to track Polymarket whale activity.

Frequently asked questions