Top-of-book
Polymarket Top-of-Book Data
Top-of-book is the venue's handshake state — the best bid and best ask with their sizes at each frame — and it is the lightest meaningful series the archive serves. This page covers what top-of-book is, what it hides, and when it is enough.
Figures measured as of 2026-09-24 on the published PolyOrderbooks archive.
Surface
The top-of-book row
Each frame carries the best bid and best ask with their resting sizes, plus the derived spread and the flags that say whether a side exists at all. On one-sided frames (16.9% of 5-minute snapshots) top-of-book reports the single side and the flag — the honest minimum.
The series is the cheapest thing you can plot that means something: direction at the executable edge, per second, for a whole market.
Limit
What top-of-book hides
Top-of-book hides depth — the sizes two levels deep that decide whether a real order fills — and it hides the bound's persistence between frames. The full ladder (ladder data) holds what top-of-book flattens, so the honest research rule is: plot top-of-book, compute on the ladder.
The measured version of the limit: top-of-book at a bound with a 42k rest and top-of-book with a 200-share rest look identical in price and opposite in consequence.
Honest
The honest framing
Top-of-book is a summary with a flag-bearing blind spot: the honest use is real-time display and quick scans, with the ladder as the canonical compute layer — the same one-grid rule the schema page states for every derived series.
Ladder
The ladder underneath
Top-of-book is a reduction of the ladder, and the ladder is what the archive ships canonically: every 250ms frame with aligned bids and asks, depth, spread, and flags. That means any top-of-book series on this site is derived, re-derivable, and auditable against the stored full L2.
The practical consequence is that research on top-of-book never has to stop at the first level — the two levels after the executable price are one query away, which is where fill outcomes actually get decided.
FAQ
What is in top-of-book data?
The best bid and ask with sizes per 250ms frame, the spread, and the side-exists flags.
What does it hide?
Depth beyond the first level — and the persistence of the bound between frames.
When is top-of-book enough?
For display and quick scans; anything decision-critical computes on the underlying ladder.