Cadence

Polymarket Cadence Explained

Cadence is the rhythm a dataset is captured at, and on Polymarket it is the deciding property: 250ms frames make reprints visible, minute grids hide them, and settlement clocks decide what a study can even ask. This page explains the venue's rhythm.

Figures measured as of 2026-09-24 on the published PolyOrderbooks archive.

Layers

Three rhythms on one clock

Why cadence matters

What cadence decides

Cadence decides which questions are answerable: at 250ms the 26-second family collapse across coins is a visible event; at minute granularity it is a step in a smoothed line. The venue's own microstructure — 90% of final-second books single-sided — is cadence-dependent by construction.

The honest consequence: a study's claim inherits its cadence, so every result on this site names the grid it used, and the latency and fidelity pages are the companion documents.

Worked example

Same event, three grids

The Aug-23 repricing event reads differently at 1s (a collapse), at 1m (a step), and at 250ms (a choreography of bid/ask rotation and reference prints). The recorded frames hold all three views; coarse data holds only the last.

A study that names its cadence can be re-verified at any grid the archive retains — the reproducibility guarantee of a multi-cadence record.

Honest

The honest framing

Cadence is a contract: "captured at 250ms" is a claim that sits between the data and every conclusion built on it, and the honest habit is checking whether a result survives a coarser re-query — the cheapest robustness test in the toolkit.

FAQ

What is capture cadence on Polymarket?

The record is captured at 250ms on the UTC grid — four full-L2 rows per second per market.

How does cadence affect research?

It decides which events are visible: the 26-second family collapse is a story at 250ms and a blink at minute granularity.

What cadence should a study use?

The finest the question needs — 250ms for settlement and reprice work — and the grid must be named with the result.