Level 2 events and snapshots
Separate research Parquet files for both contracts, grouped by CME trading day:
parquet/<contract>/<date>.events.parquet: event-driven depth updates, delivered trades and session/gap controls. No snapshot rows.parquet/<contract>/<date>.snapshots.parquet: every original captured snapshot in the day, including periodic observations. No depth or trade rows.replay/<contract>/<date>.parquet: derived viewer file, selected automatically by Inspect. It combines the observations needed for continuous replay; it is not the snapshot research table.
Choose events, snapshots or replay in the dataset subset selector. The default subset is events.
Replay rows overlap the research data and are not additional market activity.
Full captured hours, original 100 ns receipt clocks, source sequence/recording IDs and supplied timestamps are retained. Research files keep callback order, repeated trades, depth quantities and snapshot level arrays. Session controls expose action/reason; complete source descriptions remain in the original captures. Across daily files, use recording_id and sequence for source callback order; receipt clock reversals are retained. Late recovery notices retain their actual later date; a date containing only recovery controls does not imply market capture. Intervals include their start and exclude their end. Parts contain distinct data. catalog.json records file types, counts, checksums, sources and replay links.
Original captures: 2LEVEL_Data_Unverified.
These are unverified book views with up to 50 levels per side. Gaps remain gaps. Old versions are available in revision history. Provenance and continuity controls are in catalog.json.
- Downloads last month
- 75