Drift Protocol v2 (Solana) L1 Microstructure & Copy-Trading Archive (ProbioticFarmer/drift-v2-augmented-copytrading)

Georgia Tech CSE 6242 (Data & Visual Analytics) — Group 176
Harmonized Multi-DEX Market Microstructure & Copy-Trading Benchmark Archive

This dataset mirrors the exact 5-table Hive-partitioned Parquet layout and column types of ProbioticFarmer/hyperliquid-augmented-copytrading so DuckDB 2.0 (HL_DATA_MODE=hf_stream and HL_HF_REPO_ID=ProbioticFarmer/drift-v2-augmented-copytrading) can stream and join across DEXs without changing a single line of SQL:

1. perp_fills_all/date=YYYY-MM-DD/fills.parquet (29 columns): Tick-by-tick perpetual trade fills with coin, dex, price, size, side, timestamp, address, fee, is_liquidation, and liquidation_mark_px.
2. liquidations/date=YYYY-MM-DD/liquidations.parquet (29 columns): Forced liquidation fills (WHERE is_liquidation = TRUE).
3. candles_1s/date=YYYY-MM-DD/candles.parquet (14 columns): 1-second OHLCV bars (open, high, low, close, volume, volume_quote, trade_count) for ASOF JOIN execution delay shortfall (1s to 4h).
4. snapshots_perp/date=YYYY-MM-DD/<block>_<ts_ms>.parquet (12 columns): Wallet-market position snapshots (user, market, size, notional, entry_price, liquidation_price, leverage_type, leverage, funding_pnl, account_value, account_mode, date) for Pillar B1/B2 liquidation cascade & Lotka–Volterra phase-space modeling.
5. account_values/date=YYYY-MM-DD/<block>_<ts_ms>.parquet (8 columns): Daily wallet equity & long/short exposure (user, dex, collateral_token, account_value, total_long_notional, total_short_notional, account_mode, date).

Quickstart with DuckDB 2.0

INSTALL httpfs; LOAD httpfs;
SET async_threads = 64;
SET enable_http_metadata_cache = true;

SELECT coin, count(*) AS n_fills, sum(size * price) AS volume_usd
FROM 'hf://datasets/ProbioticFarmer/drift-v2-augmented-copytrading/perp_fills_all/date=*/*.parquet'
GROUP BY coin
ORDER BY volume_usd DESC;