Polymarket Ladder Data

Each 250ms frame carries aligned bids and asks — price and size per level — sorted best-first on both sides, plus depth, spread, and the flags that describe the frame. It is the only view that preserves how the venue's bound actually stood.

The derived views (top-of-book, metrics, imbalance) all reduce to the ladder, which is why the archive ships it canonically and treats every other series as a view.

Levels near the bound are the venue's gravity: executable prices bunch on the 0.99/0.995 grid 98.6% of the time, the winner side rests a median 42,254 shares at 0.99 in the 5m corpus, and one-sided frames (16.9%) show the ladder with a side missing — recorded, not repaired.

The honest research use of level data is capacity and cost: how much could trade at each level, and what walking the ladder costs — the inputs to the fill model.

Ladder data is the ground truth and the hardest to summarize: a study that reduces it to mid-means is studying a projection, and the honest habit — kept everywhere on this site — is keeping the levels in the artifacts and the summaries labeled as views.

• Depth history
• Top-of-book
• Order book data