Imbalance Labs Microstructure Sample: 7-Day Level-2 Order Book Snapshots for 24 Crypto Perpetual Futures (5-Minute, Depth-10)
Abstract
This dataset provides a free 7-day sample of time-synchronised Level-2 (L2) limitorder book snapshots for 24 cryptocurrency perpetual futures markets, resampled toa fixed 5-minute grid with no clock drift. CONTENT- 24 assets: ADA, APT, ARB, ATOM, AVAX, BNB, BTC, DOGE, DOT, ETC, ETH, FIL, INJ, LINK, LTC, NEAR, OP, SEI, SOL, SUI, TIA, WIF, XLM, XRP- 2,007 rows per asset (48,168 rows total)- Temporal coverage: 2026-02-14 23:55 UTC to 2026-02-21 23:55 UTC- Format: CSV (UTF-8, comma-separated, ISO-8601 UTC timestamps) SCHEMA (46 columns)- timestamp_utc: ISO-8601 UTC timestamp of the 5-minute bar- instrument_symbol: normalised instrument label- open_price, high_price, low_price, close_price: OHLC over the interval- interval_traded_volume: traded base-asset volume over the interval- bid_volume_level_N / ask_volume_level_N (N = 1..10): resting base-asset volume at the N-th price level of the book at snapshot time- bid_distance_level_N / ask_distance_level_N (N = 1..10): absolute price distance of the N-th level from the mid price RESEARCH USESThe dataset supports work on order flow imbalance, short-horizon price prediction,slippage and market impact estimation, liquidity and market depth analysis,bid-ask spread dynamics, execution-cost modelling, and the training and evaluationof machine-learning and reinforcement-learning agents on realistic microstructurefeatures rather than OHLCV aggregates alone. PROVENANCEDerived from the publicly available Hyperliquid L2 book archive. Raw hourly L2 booksnapshots were decompressed, parsed, aligned to a fixed 5-minute grid and reduced tothe top 10 levels per side. No interpolation or forward-filling was applied. LIMITATIONSThis is a sample intended for exploratory data analysis and pipeline validation. The7-day window covers a single market regime and is too short for robust strategybacktesting or out-of-sample model validation. FULL DATASETThe complete series covers 12+ months of continuous history for the same 24 assets atboth 1-minute (depth-30) and 5-minute resolutions:https://imbalancelabs.com/datasetsColumn-level schema documentation: https://imbalancelabs.com/schema
