Top of book
Best bid and best ask are enough for spread, mid-price and basic execution studies.
Polymarket historical order book data
Price history tells you where the market was. L2 tells you what was available.
For execution and microstructure research, best bid / ask alone can hide the depth behind the spread. BTC5m Data preserves Full-L2 snapshots and flattened levels alongside simpler quote series and quality metadata.
L1 vs L2
Choose the smallest layer that answers your question.
Full-L2 snapshots
Use depth when you need resting liquidity, book imbalance, sweeps, refill behavior or price impact.
Reference timeline
Keep external reference-price ticks aligned with the market window for cross-market research.
Replayable structure
Snapshots and levels stay tied together.
Each L2 snapshot has an identifier, outcome token, timestamps and expected bid/ask level counts. The flattened level table references that snapshot, so validation can verify that the number of levels written matches the snapshot metadata.
This is more useful for reproducible research than a visual book screenshot or a few sampled prices.
- Snapshot key
- snapshot_id + token
- Sides
- bid / ask
- Level order
- level_index
- Values
- price + size
- Validation
- level-count parity
What “historical” should include
Data quality needs evidence, not just a file name.
✓Window start and end are explicit
✓Source timestamps and receive timestamps are separate
✓Connection gaps and missing-rate are reported
✓Every packaged file has a checksum
✓Partial captures stay rejected
Looking for a specific market period?
Request the date range before we build more product.
We use real requests to decide which periods and fields deserve packaging next.