Imbalance data
Polymarket Order Book Imbalance Data
Order book imbalance — the ratio of resting size on one side to the other — is Polymarket's barometer of one-sidedness, and near settlement it is the whole story: 76.2% of final-minute snapshots in 5-minute markets have an empty side. This page covers the imbalance series as a data product.
Figures measured as of 2026-09-24 on the published PolyOrderbooks archive.
Coverage
The imbalance series
The archive computes imbalance as top-of-book asymmetry — bid size over total depth at the first levels — and stores it per row on the same 250ms grid as prices and metrics, so any study can reproduce it from the raw ladder rather than trust a derived number.
Imbalance is the venue's most informative single quantity: its extremes are the 16.9% one-sided snapshots and 3.24% crossed states that define Polymarket's liquidity regime.
Measured
What measured imbalance shows
In the 5-minute corpus, imbalance is not noise: 90% of final-second books are single-sided, the winner side rests a median 42,254 shares, and executable prices sit at the 0.99 bound rather than at an interior price.
The measured lesson is a data-product one: imbalance series let a strategy detect when the book has made its decision and when it still has two sides — the difference between joining and chasing at settlement.
How to use
Using the series
Pull the imbalance column for a slug; it moves in steps as depth drops off, and its relationship to the executable price is the tradable signal — extreme imbalance with price at the bound means the market is paying to take one-way liquidity.
For studies, keep the crossed and one-sided flags: they and the imbalance column tell the same story in two notations.
Access
Getting imbalance data
Honest
The honest framing
Imbalance is descriptive, not predictive: the archive measures it faithfully and the honest use is as a regime detector and a fill-cost input, not as a standalone forecast feature.
FAQ
What is order book imbalance on Polymarket?
The asymmetry of resting depth between bid and ask sides, stored per 250ms row and recomputable from the raw ladder.
Is imbalance strong near settlement?
Yes — 90% of final-second books in the 5-minute corpus are single-sided; imbalance saturates as one side empties.
How do I get the series?
Export per-row imbalance with prices and metrics in CSV/JSON/Parquet, or recompute it from the recorded full-L2 ladder.