diff --git a/README.md b/README.md
index 7abc0e5..eae5de8 100644
--- a/README.md
+++ b/README.md
@@ -238,6 +238,12 @@ let candles = info_client
 // Supported coins: BTC, ETH, SOL, AVAX, MATIC, ATOM, and many more
 ```
 
+### Bulk Historical Data (Alternative Source)
+
+The examples above pull candles from the Hyperliquid info API. If you need longer or more granular history for backtests, such as trade-level fills or order book state, pre-built Parquet datasets are another option:
+
+- **[Hydromancer Reservoir](https://hydromancer.xyz/hyperliquid-historical-data)**: Hyperliquid historical data on AWS S3 (`s3://hydromancer-reservoir`, region `ap-northeast-1`, requester pays), stored as Parquet and partitioned by `date=YYYY-MM-DD`. It includes trade fills (with liquidation, ADL and TWAP subsets), 1-second OHLCV candles, 20-level order book snapshots at 1-minute cadence, and daily position snapshots, for Hyperliquid perps and HIP-3 dexes. Any Parquet reader (DuckDB, Polars, or the `arrow`/`parquet` crates) can load it into the OHLC vectors used in the backtesting example above. It does not include funding rates, so keep using the API for those.
+
 ### Trading Modes
 
 The library supports multiple trading modes through the unified interface:
@@ -574,4 +580,4 @@ This software is for educational and research purposes only. Trading cryptocurre
 
 - **Issues**: [GitHub Issues](https://github.com/xsa-dev/hyperliquid-backtest/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/xsa-dev/hyperliquid-backtest/discussions)
-- **Documentation**: [docs.rs](https://docs.rs/hyperliquid-backtest)
\ No newline at end of file
+- **Documentation**: [docs.rs](https://docs.rs/hyperliquid-backtest)