Polymarket Order Book Depth, Explained

• Top-of-book depth: contracts at the best bid and best ask — what you can hit first.
• Ladder depth: contracts at each level below the top, in ticks (35–36 levels per side measured on sports).
• Depth imbalance: bid-to-ask size ratio at the inside levels.
• Crossed depth: levels where a bid exceeds an ask, marking fast repricing.

The measured record is consistent: in the August 23 event, five crypto markets hit lower bounds within 26 seconds while depth was leaving and inverting ahead of the prints; in the ETH study, bid depth fell 96% and ask depth rose almost sixfold before the price caught up. Depth is the conference, price is the press release.

At settlement the same pattern appears at scale: the winning side accumulates 42,000+ shares at a fixed 0.99 while the losing book empties. The depth story is decided before the resolution print.

• Compute usable depth: size at your price level on your side — the realistic fill size.
• Track imbalance thresholds as long-horizon signals; the archive shows them persisting longer than price moves.
• Replay fills against the recorded ladder instead of the mid, so partial fills and slippage are modeled, not assumed.
• Detect inversions by flagging crossed rows, which are preserved in the data as the defect they are.

• For one market, pull the full frame list at 250ms for an hour and count the distinct levels per side (sports: 35–36; Up/Down: variable) — depth begins as a level count.
• Compute per-frame top-of-book sizes and their median; compare to the median trade size (42,254); the ratio shows how far one trade reaches into the book.
• On a reprice window, measure per-side depth change; ETH's measured −96% bid / +491% ask on August 23 is the canonical depth redistribution to reproduce.
• Label each frame as two-sided or one-sided; a depth claim that ignores one-sidedness is a claim about half the book.

Depth is the multi-level face of the book: the level count defines the ladder, the sizes define reach, and the one-sided share defines reactivity — the three together, not any one, describe Polymarket depth.

The measured depth behavior — sports ladders steady at 35–36 levels per side, Up/Down ladders compressible to a single level near resolution, and depth that redistributes violently during repricings — is the venue's actual depth profile, and the archive is where it becomes a query.

A depth profile is three columns: levels per side, median top size, and the one-sided share. Compute them per market, print them side by side, and the nine-in-ten one-sided reality stops being a claim and becomes a table.

The depth summary worth publishing is therefore small and dense: one-sided share, median top-of-book size per side, and the executable side label, each computed from a stated window. That table reproduces from raw frames in one query and makes every later claim about liquidity checkable against the same numbers.

Depth is the book's three-dimensional view: levels, sizes, and sides.

One-sidedness makes per-side depth the only honest measure; aggregate "depth" hides the venue's defining trait.

Reprice windows are where depth becomes flow — the −96/+491 swing is the same event from two sides, and the archive stores both.

Depth is the sizes layer of the book: levels, sizes, and sides all stay distinct in this page, which makes it the reference for the liquidity page.

It backs the depth-data page and the market-making page; both cite the measured per-side redistribution during reprises.

The one-sided share is the column that connects depth work to the venue's defining trait.

Depth is levels, sizes, and sides; one-sidedness is the third of those and the easiest to forget.

Measure per-side depth and never let an aggregate number hide the venue's defining trait.

• Order book data
• Price history
• Crypto data
• Provider matrix