Market LOB (L50, ITCH-like) — Synthetic Message-Level Order Book + Calibration Card (Premium Eval Kit)

What this is (in one line)

Why this is sought-after

• Exchange-grade fidelity: Message-level events at ns precision with queue-aware levels and periodic depth snapshots (L1–L50).
• Plug-and-play: ITCH-like schema and Parquet-first delivery—works immediately with Pandas/Polars/Spark and most backtest stacks.
• Receipts included: Ships with a Calibration Card and QC pass/fail gates (monotonic timestamps, non-negative spread/depth, cancel/amend rate bands).
• Research-ready: Perfect for rapid strategy prototyping, execution algo tuning (VWAP/TWAP/POV), risk stress tests, and model validation—without touching real exchange feeds.
• Procurement-friendly: 100% synthetic by construction (no real counterparties), easing compliance and onboarding.

What you get inside the ZIP

• data/<SYMBOL>_<DAY>_L50_events.parquet — ITCH-like message stream: ts_ns, msg_type, order_id, side, level, px, qty, trade_qty, trade_fills
• data/<SYMBOL>_<DAY>_L50_snapshots.parquet — Depth snapshots (L1–L50) with derived best_bid, best_ask, spread plus columns bid_qty_Li / ask_qty_Li for i=1..50 (CSV previews included for both.)

• docs/schema.json — Field dictionary & data types
• docs/manifest.json — Scenario summary (depth, tick, duration, seed, notes)
• metrics/calibration_card.json — QC results + distributional receipts
• samples/* — Preview CSVs you can open in seconds
• checksums.sha256 — Integrity verification

Typical contents (Premium Eval Kit)

• Scope: 1 synthetic “day” (configurable duration) at L50 depth
• Events: NEW, AMEND, CANCEL, TRADE, SNAPSHOT
• Timebase: integer nanoseconds since epoch
• Schema parity: ITCH-like (house converter to full OUCH/ITCH mapping available on request)

Who should buy this

• Quant researchers / HFT & execution teams: test order placement, queue dynamics, impact & resiliency without revealing intent.
• Risk & microstructure teams: run volatility/cancellation stress tests on safe synthetic tapes.
• Data science leads: validate pipelines, feature engineering, labeling, and evaluation notebooks—fast.

Quality & Calibration (what we guarantee in this kit)

• QC gates:
  • Timestamp monotonicity (events & snapshots) — PASS
  • Non-negative spread and depth — PASS
  • Cancel/Amend rate bands within realistic ranges — PASS
• Calibration Card includes (kit-level):
  • Depth L1 means (bid/ask), inter-arrival distribution KS stats, trade quantity ACF excerpt
  • Notes on scenario parameters (tick size, event rate, seed)

Need strict ITCH/OUCH field names or a Counterfactual Diff Report (baseline vs shocks)? We can provide a converter + reports on request or via enterprise bundles.

License & compliance

• Evaluation License (90-day, non-production) for internal research and benchmarking.
• 100% synthetic by construction; no PII/PHI; not derived from any single real exchange participant’s data.

How to load (Python)

import pandas as pd
events = pd.read_parquet("data/<SYMBOL>_<DAY>_L50_events.parquet")
snaps  = pd.read_parquet("data/<SYMBOL>_<DAY>_L50_snapshots.parquet")
print(events.head(), snaps.head())

Support

• Response: within 1 business day
• Fixes: material schema/data issues within 5 business days
• Contact: joshua@alita-therapeutics.com

Looking for more? We offer enterprise day/week packs, counterfactual shocks, exclusivity, and subscriptions (private listing / direct PO). Ask for the Enterprise LOB Catalog.