Polymarket settlement windows: 5m, 15m and 4h

• 5,005 crypto Up/Down markets in the archive: 3,467 at 5 minutes, 1,436 at 15 minutes and 102 at 4 hours
• 5-minute markets are 69% of the total; 4-hour markets are about 2%
• Every horizon settles on the same 60-second Chainlink TWAP (twap_lookback_seconds = 60)
• Bitcoin and Ethereum together account for 31% of all crypto Up/Down markets

Measured on 2026-10-17, the archive holds 5,005 crypto Up/Down markets split across three horizons: 3,467 five-minute, 1,436 fifteen-minute and 102 four-hour contracts. The distribution is heavily front-loaded - short windows are where the volume and the research interest are.

That skew is not uniform across assets. Bitcoin alone contributes 598 five-minute markets, while the smallest asset, Zcash, contributes 202 across all horizons combined. A study that pools assets without weighting is really a study of Bitcoin and Ethereum with some noise attached.

It is tempting to assume a 4-hour market settles on a 4-hour average. It does not. Every market, short or long, settles on a 60-second TWAP - twap_lookback_seconds is 60 across the board. The horizon sets how long the contract lives; the TWAP window sets how the settlement price is smoothed.

This has a concrete consequence for backtesting. A 4-hour market's outcome depends mostly on where the asset sits in the final minute, not on the path across four hours, so a strategy that predicts the path but not the endpoint will still miss. Modeling the close correctly matters far more than modeling the middle.

Short windows produce a new market every five minutes, so a single day yields hundreds of samples per asset. That density is why 5-minute markets are the natural unit for statistical studies: within one trading day you can collect a large sample of near-identical contracts.

The 4-hour market is the opposite. With only 102 across all eight assets in the whole archive, there is not enough data for most analyses; treat any 4-hour result as a case study rather than a distribution.

For a balanced study, decide the unit before you pull data. Combining 5-minute and 15-minute markets weights the sample toward what traded most, which is fine if that is your question and misleading if it is not.

Because all three horizons share the same 60-second settlement rule, you can compare them cleanly on spread, depth and cross rate without worrying that the settlement methodology differs. The differences you measure are market-structure differences, not methodology artifacts.

import requests

BASE = "https://api.polyorderbooks.com/v1"
HEADERS = {"X-API-Key": "your-key"}

# Enumerate BTC markets and bucket them by horizon from the slug.
cursor = None
counts = {"5m": 0, "15m": 0, "4h": 0}
while True:
    params = {"event_slug": "crypto-up-down", "limit": 500,
              "category": "crypto"}
    if cursor:
        params["cursor"] = cursor
    r = requests.get(f"{BASE}/markets", headers=HEADERS, params=params)
    r.raise_for_status()
    body = r.json()
    for m in body["data"]:
        for h in counts:
            if f"-{h}-" in m["slug"]:
                counts[h] += 1
    cursor = body.get("metadata", {}).get("next_cursor")
    if not cursor:
        break
print(counts)

The Starter plan is free, includes order books, prices, and metrics at 250ms resolution, with 3 days of history, 60 requests/min, 1,000/day, 1 free AI backtest, and 3 strategy backtests. No credit card required.

Paid data windows from $19/mo extend history to 30–120 days at higher throughput. Backtest AI add-on is +$19/mo or standalone at $29/mo.

• 5-minute markets
• Crypto data overview
• Chainlink TWAP resolution