BTC 15m Observatory
Polymarket captured data
Connecting

Polymarket BTC 15-minute markets

Every UP/DOWN market is recorded from the public order-book feed and kept as replayable Level-2 depth. Browse what happened, or test what a set of trading rules would have done against the same recorded books.

History

Pick any recorded 15-minute market and step through it: UP and DOWN price paths, spreads, the full reconstructed order book at any instant, public trades, and the BTC reference feed.

Browse recorded markets →

Backtest

Set an entry price, an optional time window, an optional stop loss, and a fixed or compounding stake, then run it across every recorded market. Orders are filled by walking the real order book, so size that the book could not absorb is reported as unfilled.

Configure a strategy →

Read-only. This site holds no wallet, submits no orders, and never writes to the recorded archive. Backtests are simulations over past order books; a simulated fill is an estimate of what the book would have given you, not a promise. Past results do not predict future ones.

Backtest a strategy

Buy a side once it reaches a price, then hold to settlement — or sell if it falls back to a stop loss, which you can arm only from a chosen minute. Every order is executed against the recorded Level-2 book with latency, so it can fill partially or not at all.

Parameters
Loading recorded coverage…
Measured on the ask — the price you would actually pay, not the midpoint. This is the trigger, not necessarily the price you get.
The highest ask the order will accept. It matters more than it looks, because the trigger above only says when to buy, not at what price:
  • If the ask is already past your trigger when the window opens — normal when the window starts late — you buy at whatever it is now, up to this ceiling.
  • If the book moves while your order is in flight, anything above this is abandoned instead of filled.
A binary bought at p has to settle in your favour p of the time just to break even, so a loose ceiling quietly raises the bar. Keep it close to the trigger — 0.81 for a 0.80 trigger — unless you deliberately want to chase.
Which side
“Favoured” buys whichever side reaches the entry price first.
Staking
The same stake on every market, so one bad market cannot end the run.
Converted to shares at the fill price. If the book is thinner than your size, the remainder simply does not fill — the result reports how much.
Timed entry
Off means the entry may trigger at any point in the 15 minutes.
Stop loss
Sell everything if the bid falls to this price. Off means hold to settlement.
Arm the stop later
The stop does nothing before this minute, so an early dip is ridden out and only a late collapse is sold into. Off means the stop is live from the moment you are filled.
How much of the displayed book you are assumed to actually get, how far your order will chase it, and how long it takes to arrive. A result that only works under “optimistic” has not been demonstrated — try all three.
How the parent entry is split after the signal. Adaptive POV reacts only to volume already observed, keeps a partial time-based floor, and catches up at the deadline. TWAP follows an even schedule. Immediate sends the full affordable order at once.
The deadline for adaptive POV or TWAP child orders, capped by the entry window and market close. Immediate execution does not use this horizon.
Markets to test
Leave empty to use every recorded market.
No backtest has been run yet Choose parameters and press Run. Every selected market is replayed order-book state by order-book state, so a run takes a few seconds.
Simulated fills over recorded public order books · no wallet · no order submission

About this platform

This site records every Polymarket BTC UP/DOWN 15-minute market straight off the public CLOB WebSocket feed — full Level-2 order books, trades, and the Binance/Chainlink BTC reference price — and normalizes the raw messages into structured, decimal-clean datasets. History replays that data as price charts and reconstructed order books; Backtest replays it as fills, running a strategy against the same recorded books. Nothing here is a wallet, and nothing here submits an order.

Captured trade volume—Executed notional, all recorded markets
Markets recorded—15-minute UP/DOWN intervals
Trades captured—Public fills on either side
How it works
Raw feed → normalized data → charts and backtests
01

Collected live over WebSocket

A collector subscribes to the public CLOB market-data socket for every BTC UP/DOWN market the moment it opens, and keeps the full, unfiltered stream — snapshots, level updates, and trades — with nothing summarized away.

02

Normalized to fixed-point, decimal-clean data

Raw messages are parsed into typed tables: order-book snapshots, level updates, trades, and BTC reference prices, all at exact fixed-point scale so nothing is silently rounded through floating point.

03

Visualized as history

The History view turns that normalized data into UP/DOWN price paths, spreads, and a full reconstructed order book at any instant you pick.

04

Replayable for backtesting

The Backtest view runs a strategy — entry price, stop loss, position size, staking — against every recorded market by walking the real recorded book, so fills reflect what the book could actually have given you.

Download the normalized data
CSV columns: slug, condition_id, outcome, t_ms, iso_time, best_bid, best_ask, midpoint, spread
The exact slug shown in History, e.g. via its "Copy slug" button.
Every market starting in this window, combined into one CSV. Leave one side blank to leave that edge open — but set at least one.
One row per top-of-book observation across every market. This can be a large file and take a while to generate as the archive grows.

Read-only. Downloads stream straight from the same normalized archive the charts read — no wallet, no credentials, no order submission anywhere on this site.