Polymarket Candlestick Charts, Explained

• OHLC derives from executions on each token in the window.
• Open = first fill price, close = last, high/low = extremes of the fills, volume = shares traded.
• Candles exist at many resolutions (1s to 1d per provider); the more fills, the more faithful the candle.
• Trades settle on-chain, so candles are rebuildable from the chain — verifiable.

Candles hide the book: the depth behind the close, the spread during the window, and whether the close was executable. A candle can close at 0.50 that printed across a 99.6-cent chasm — the OHLC is true and the trade was trapped.

For backtests, candle-based fills overstate realism because fills are assumed at OHLC prices that rarely existed with size.

Pairing candles with book metrics (mid, depth, spread per window) is how this archive keeps both views without choosing.

• Match resolution to horizon: 1s–1m candles suit five-minute Up/Down; 4h+ candles for long research.
• Use trade-volume from fills, and separately read book depth for the quote side.
• Verify with on-chain fill counts when the question is provenance, not convenience.
• For terminal windows (final minute), prefer book rows over candles — settlement is a depth story.

• Derive a minute candle from the 250ms frames for the hour you studied: open at boundary, high/low inside, close at boundary, volume summed.
• Verify one candle against its frames (the round-trip query) and note any second whose intra-candle range a candle would hide; the 99.6-cent second is the extreme example.
• Read the same hour as 1m, 5m, and 15m candles and record how the visible structure changes; granularity choice is structure choice.
• Always store the candle-relevant resolution in the file name; a candle without its resolution is a chart without a scale.

Candles are a windowing decision, not a truth: open/high/low/close/volume summarize frames, and the summary hides whatever crosses between boundaries — the venue's fastest moves live at 250ms and are invisible in a 15m candle.

The failure mode is not the schema but the habit: a "quiet" candle can contain a one-second full-ladder pass, so candle work has its value only when it promises the underlying frame it came from.

The descent rule for every chart reader: any candle worth acting on gets opened to its frames, and any frame that surprises gets opened to the second around it. That three-level descent is the entire candlestick discipline, and it costs one query per step.

A bare minimum for candle labeling, then: window boundary (start and end UTC), first and last executable prices — not mids — and the volume of the window in contracts. Anything less is a digest, not a chart, and digests are for navigation, not conclusions.

The 15-minute and four-hour pages in the data family reuse this exact labeling, so a reader who masters the descent rule on one page can apply it to the rest of the archive unchanged.

Candles serve the digest; the frames serve the truth.

Every candle you publish should carry its resolution and, when it matters, the frame it came from.

The ninety-nine-cent pass teaches the rule: never let a compressed summary outrank the frames it summarizes.

Candles are the digest layer built over the frames; this page explains when to trust them and when to descend to 250ms.

Pair it with the OHLC and 15m/4h data pages for the row-level view of the same aggregation.

The "compression hides the 99.6-cent second" warning is the rule every chart-reading page inherits.

Candles are a windowing decision over frames, valuable for digests and dangerous for discovery.

Never let a compressed summary outrank its frames; the 99.6-cent second is the rule's proof.

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