90-Day ETH 15m Polymarket Data Download
Download 90 days of ETH 15m Polymarket data for local analysis, Parquet workflows, research, and reproducible strategy tests.
90-day ETH 15m Polymarket data download gives traders and researchers a rolling multi-month sample of Ethereum Up/Down markets. PolyHistorical captures full-depth order book snapshots at 300ms granularity, so data export can account for spread, liquidity, and the prices that were available at each point in a market.
A complete 90-day calendar window can contain up to roughly 8,640 scheduled 15m windows before launch dates, feed gaps, or unavailable markets are considered. Always measure the returned coverage instead of assuming every possible window exists.
Why Use a 90-Day Window?
Fifteen-minute contracts balance sample size with enough time for liquidity, momentum, and mean reversion to develop. A 90-day lookback is long enough to compare multiple volatility and liquidity regimes while remaining practical for rolling research and repeated model validation.
Build a Reproducible 90-Day Download
For a local ETH 15m research set, export the raw snapshots before calculating features. Partition files by market date and identifier so a failed job can resume cleanly and individual contracts can be audited later.
- Keep raw prices and sizes separate from derived midpoint or imbalance columns.
- Use UTC timestamps and preserve the original market and token identifiers.
- Write columnar files for scans, plus a small manifest with row counts and coverage.
- Record the retrieval date and filters alongside every export.
Quality Checks Before Analysis
Historical depth is only useful when the sample is auditable. Check that timestamps are ordered, both outcome tokens are present, price levels are valid, and market resolution is known. Flag snapshot gaps instead of forward-filling across them, because invented book states can create fills that never existed.
For strategy work, avoid using midpoint as an execution price. A midpoint is descriptive; an order must interact with the bid or ask and may consume several levels when the requested size is larger than top-of-book liquidity.
90 Days vs 120 Days
Choose 90 days when that window matches your subscription, compute budget, or validation schedule. The 120-day ETH 15m data download page covers the longer alternative. Using both as rolling horizons can show whether a result is stable or overly dependent on the most recent regime.