Polymarket OHLC Data

• Minute candles: 1,440 rows per day per market — the intraday trading frame.
• Hour candles: 24 rows per day — the session frame.
• 4h/day candles: 6/1 rows per day — the digest frame.
• Each candle carries open, high, low, close, and volume at the same interval grid; none of the prices are exotic — they derive from the ladder stream.

A high-to-low range of a cent looks like quiet; the same two frames can contain a 99.6-cent ladder in one second (observed 0.044 to 0.996), which a candle hides by construction.

The archive lets you validate any claim about candles by pulling the underlying frames for the interval — the honest check for every chart you publish.

Volume-at-frame plus range-at-frame means "candle with participation" is a computed column, not a separate dataset.

• Screening hundreds of markets by daily range and volume is a candle query; deep-diving one market is a frame query.
• Execution studies benchmark fills against candle open, while frame studies benchmark against the executable floor — use each for its kind.
• Digest reports read 4h/day candles; research reads 250ms frames; the pipeline between is one aggregate function.

• Pull minute candles for one active sports market and one Up/Down market; count rows — expect 1,440 per day per market.
• For the highest-range candle of the day, open the same window at 250ms and check the frame actually moved inside it; the candle range should trace to real frames.
• Compute per-day volume-at-range: daily aggregate volume divided by the day's candle range, a single-number activity metric per market.
• Save day candles for a full series and list the top range days — the schedule dates that mattered.

Candles are a data product, not a data source: they summarize a frame record, and the quality of the summary depends on the window math, which is why the archive lets you verify every candle against its frames.

The failure mode is compression: a one-second ladder reprice (0.044 to 0.996 observed) inside a "quiet" minute candle is real movement that the candle hides by its own definition.

Any claim about high/低/close should carry the frame it came from; that is the discipline that separates chart summaries from tradeable statements.

Candles serve screens and digests; frames serve the moment. Keep the candle layer as a derived cache — overnight jobs write it, dashboards read it, and studies skip it.

When a candle and a frame disagree, the frame wins; build that precedence into your reporting and your tooling.

Candles are the compression layer, and compression has a price: the one-second 99.6-cent ladder pass is invisible in a minute candle even though it is the most important movement of the hour.

Every candle claim in your reports should trace to frames: open/close at boundaries, high/low inside, volume summed — the round-trip check is one query and it is the difference between a chart and a hypothesis.

Let candles run the screens and digests, and let the frames run the decisions; a dashboard that shows minute bars with a jump-to-250ms control is the exact design this page recommends.

When a candle and a frame disagree, the frame wins; encoding that precedence into your tooling prevents the subtle errors that make compressed views convincing but wrong.

The honest candle workflow is: derive candles from frames, validate one candle against its frames per run, and never let a candle claim a time-precision the frames deny.

A minute candle is the right unit for flow-level work; the 99.6-cent one-second ladder pass belongs to the frames page, and the OHLC page should send you there when your range is that violent.

Volume-at-range is the derived metric most reports underserve: sum the window volume, divide by the range, and you have an activity-per-unit-move number that survives normalizing across markets.

The rule that keeps the layer honest: candle wins for screens, frames win for decisions, and the precedence is encoded in tooling — when they disagree, the frames are the truth.

• candlestick charts explained
• vwap data
• 4h data