← Back to list

Polymarket has no historical order book. We built one.

If you have ever tried to backtest a Polymarket strategy, you already know the problem. You can see the current price. You cannot see what…

Aditya Bajaj · 2026-07-01 12:56 · 0 claps · 2.3 min read
#polymarket #prediction-markets #cryptomarket #quant #data
Open on Medium ↗
Wiki topics: CRY · Crypto & Web3 BIZ · Business Strategy ECO · Economy · General HIS · History

Polymarket has no historical order book. We built one.

If you have ever tried to backtest a Polymarket strategy, you already know the problem. You can see the current price. You cannot see what the order book looked like five minutes ago, five hours ago, or five days ago. Polymarket does not publish that data. It never has.

That means every “strategy” traders talk about on Twitter is a guess dressed up as an insight. No depth history. No way to check if a price move was backed by real liquidity or a single thin order getting picked off. No way to test if your idea would have actually worked before you put money behind it.

We got tired of that gap, so we built the dataset that should have existed already.

What that actually means

Resolved Markets has been capturing full order book depth, tick by tick, at millisecond resolution, since we started. Not just the best bid and ask. The entire depth array, both sides, every change. Right now that is over 700 million snapshots across more than 100 markets, spanning crypto, sports, economics, weather, social, equities, and Hyperliquid perpetual futures.

If a market moved, we caught it. If liquidity dried up right before a resolution, we have the timestamp. If a large order sat on the book for three seconds before getting filled, that is in the data too.

Who this is actually for

Three types of people have been asking us for this, consistently:

Polymarket traders who want to know if their edge is real before risking capital on it.

Crypto quants who are used to having this kind of data for every exchange and were baffled that prediction markets had nothing comparable.

People building AI trading agents who need clean, structured, high resolution market data to feed a model, not scraped screenshots of a price chart.

How you get at it

We did not want to build one interface and call it done, because these three groups work completely differently. So there is a REST API for people who want to pull historical data directly. A WebSocket stream for anyone tracking markets live. An MCP server so you can plug this straight into Claude or another AI agent and ask it questions about market history in plain language, full setup guide at resolvedmarkets.com/docs/mcp. A no-code Visual Strategy Builder for people who want to test an idea without writing a query. And an AI Backtest Agent for people who want to describe a strategy and get a result back.

Why this matters more than it sounds

Data resolution is not a minor technical detail. It is the difference between knowing what actually happened in a market and guessing based on what the chart looks like after the fact. A price chart smooths everything out. An order book does not lie. If you are serious about testing a strategy instead of hoping it works, you need the second one, not the first.

We are not claiming every backtest run on this data will make money. Most will not. That is the point of a backtest. But right now, most people trading Polymarket cannot even run one, because the data to do it has never existed publicly until now.

Try it at resolvedmarkets.com


메타데이터
post_id
a2cedb4ef025
slug
polymarket-has-no-historical-order-book-we-built-one-a2cedb4ef025
url
https://medium.com/@AdiFlips/polymarket-has-no-historical-order-book-we-built-one-a2cedb4ef025
canonical_url
https://medium.com/@AdiFlips/polymarket-has-no-historical-order-book-we-built-one-a2cedb4ef025
author_url
https://medium.com/@AdiFlips
status
ok
fetched_at
2026-07-10 17:29:03