Polymarket election data: where it lives, what's free, and the depth gap

Election queries are among the most-searched on Polymarket, and the data behind them is genuinely free. Gamma serves election metadata, outcomes, resolution, volume, and liquidity — including markets resolved years ago — and CLOB prices-history serves per-token price series back to market creation at a 1-minute floor. Neither needs a key.

Resolution and price history are time-ordered, so election studies mostly hang on prices-history behavior — and that has a documented edge case, measured rather than assumed.

Bound a request with startTs and endTs and 1-minute buckets come back normally: a 2024 election market resolved ~21 months earlier returned 360 price points for a 6-hour window at fidelity=1. Unbounded, the same market returned 0 points at fidelity=60 and 615 points at fidelity=720. The empty responses come from interval=max, not from resolution (py-clob-client issue #216).

What free election data does not cover is depth. The CLOB /book returns the current ladder only; there is no historical book endpoint and no archive behind it. For the 2026 cycle this matters because an election book is only worth studying once you have stored it. Our archive captures full L2 ladders every 250ms for crypto up/down and sports, not election contracts — so we say plainly: election depth history is forward-only.

• Gamma API — election metadata, outcomes, resolution, volume, and liquidity with full history, free (what the API serves).
• CLOB prices-history — election token prices at 1-minute buckets, free; bound the request with startTs and endTs (price history).
• Resolution checks — settled election markets keep working after resolution; the pricing page measures this behavior (API pricing).
• Running your own collector — for election L2 depth, capture it live; the order book data guide covers the schema and resolution to record.

Election data on Polymarket sits behind the official free APIs: Gamma for metadata, outcomes, volume, and liquidity, and CLOB prices-history at a 1-minute floor. For live questions — what does the favourite trade at now — that stack is sufficient and costs nothing.

The measured behavior of prices-history on a resolved election market is the education: after resolution the series freezes to the terminal answer, so any backtest built on the 1-minute floor inherits its fidelity limits, and any depth question has no official answer at all.

That is the depth gap: a ballot-trading study needs the ladder as it actually stood — the one-sided books, the bound-side executable prices, the second-by-second reprise on a poll release — and none of that exists in the free endpoints. It exists only where someone recorded the whole book, and recorded books are what the archive does.

The honest template for election research uses both layers: the official API for the market universe and live prices, and the stored 250ms archive for any historical depth the study will cite. The two join on the UTC second, exactly as the venue's own price and book timestamps.

Name the resolution mechanism before quoting any election price: a single traded outcome after resolution is a settled fact, not a live signal, and mixing the two corrupts a backtest.

The favourite's bound-side resting depth — often a large block at 0.99 — is the election microsignal; it says where the crowd is parked, and it only exists in recorded depth.

Watch poll-release repricing in the ladder, not the 1-minute series: a minute grid flattens a two-second reprise into a single stair that looks gradual.

For any number a study will publish, keep the stored frames; the archive is the citation that lets a reader re-run rather than retype.

• What the API serves
• Price history
• API pricing
• Order book data