Polymarket 4-Hour Data

• Multi-week arcs: a two-week tournament at 4h is 84 rows per market — a readable line for arc analytics.
• Cross-market comparison at scale: scan dozens of markets on spread/volume at 4h without frame noise.
• Aggregated liquidity trend: top-of-book depth at 4h intervals reflects the liquidity regime, not its micro-moves.
• Long-horizon strategy baselines: fast variables measured at 4h are the settlement-school analog of a weekly look.

Nothing about a reprice exists at 4h resolution: the 65-second ETH repricing, the 26-second coin collapse, and final-minute ask absence are all invisible.

What 4h preserves is the continuity of state: whether a market traded above/below a level in a window, whether volume collapsed, whether the mid held a bound.

Use 4h to set expectations and 250ms to verify them; every 4h claim should be reproducible against the frame record it summarizes.

• Open/high/low/close/volume per 4h window, aggregate of the 250ms frames beneath.
• Same histogram of fields as every other cadence; the schema does not change shape with horizon.
• Scheduling reports at 4h keeps notices quiet (6/day/market) while politicking stays unseen.

• Pull a two-week sports series at 4h: 6 rows/day = 84 rows per market, and the arc fits on one screen.
• Build the volume-at-regime metric: sum the 4h volumes across the day and divide by the day's high-low range — activity per unit of movement.
• Scan a list of markets at 4h for open-interest direction changes; the OI trend is the slowest honest signal the cadence preserves.
• For one 4h window, verify its open/close against the 250ms frame at the boundaries; the window math must round-trip.

The 4h view is the horizon scanner: six rows a day keep the record continuous without pretending to contain the second-level behavior that drives those six rows.

Its strength is exactly what it excludes: drawing a two-week market line at 4h is readable in a way that 250ms charts never are, and the structure it shows is real, windowed truth.

Keep it honest by the same rule as 15m: any claim it makes about a specific time is a claim about the frames, and the frames back it or not.

4h is the scheduled-report cadence: six quiet notices a day, one readable digest, and a scan table for long horizons.

Anchor it in storage next to the raw record so re-derivation is a refresh, not a re-purchase.

The 4h view is the horizon scanner: six rows a day give a two-week series in one readable line, and open-interest direction plus volume-at-regime are the slowest honest signals that survive the cadence.

Respect what the cadence excludes the same way you respect what it keeps: a 65-second repricing and a 26-second family collapse cannot appear at 4h, and the page says so rather than pretending otherwise.

Window math is the discipline: each 4h row's bounds and volume must round-trip through the frames, and horizon-aligned filters belong in every participant-facing report.

Schedules and digests read 4h; decisions read 250ms. The two are not rivals — one sets expectations, the other verifies them, and the archive provides both from a single source of truth.

The 4h lens is for the longest view: a two-week series as 84 rows, open-interest direction as a slow honest signal, and volume-at-regime as the activity benchmark that survives the horizon.

The boundary is the same as every coarser cadence, spelled out: the 65-second repricing and the 26-second family collapse require 250ms, and the page refuses to pretend to contain them.

Use the window round-trip as the trust contract: each 4h row's open/close/high/low/volume must reconstruct from the frames; when they do, the digest and the archive are the same story.

Scheduled reports that read 4h and decisions that read 250ms are the pair this page recommends — different grain, one source of truth, and no number that cannot be traced back to a frame.

• 15m data
• ohlc data
• intraday data