0xArchive comparison · Backtesting
A market data API is good for backtesting when it has the history your strategy needs, at the grain your fills need, with gaps you can see before they skew a result. For Hyperliquid and Lighter that means four checks: when the book history starts, whether the book is price levels or single orders, whether funding and liquidations are stored, and whether the provider tells you where data is missing.
0xArchive is built around those checks for Hyperliquid and Lighter. Provider facts below come from each provider's own documentation, checked on 24 October 2026.
What a perp backtest needs from its data
| You need | Why it matters | 0xArchive |
|---|---|---|
| Book history | Fill prices depend on depth, not the last trade | Hyperliquid from 15 Apr 2023, Lighter from 29 Jan 2026 |
| Single orders in the book | Queue position decides whether a limit order fills | Hyperliquid L4 from 10 Mar 2026, Lighter L3 from 5 Mar 2026 |
| Complete trades | Volume, aggressor side and slippage checks | Hyperliquid from 22 Mar 2025, Lighter from 17 Jan 2025 |
| Funding history | Carry decides whether a held perp position pays | Hyperliquid from 20 May 2023 |
| Liquidations | Cascades move price faster than candles show | Hyperliquid from 22 Dec 2025 |
| Known gaps | A missing hour reads as a quiet hour | Coverage, gap and incident routes |
Book history is where Hyperliquid providers differ most
| Provider | Hyperliquid history from | Stored order-level book |
|---|---|---|
| 0xArchive | Book 15 Apr 2023 | From 10 Mar 2026 |
| CoinAPI | 27 Sep 2024 | Hourly files from 9 Jun 2026 |
| Tardis.dev | 29 Oct 2024 | None, price levels only |
| Dwellir | Node archive 25 Jan 2025 | Raw diffs from 22 Jan 2026 |
| GoldRush | 25 May 2025 | Not stated |
| Hydromancer | Free archive 28 Jul 2025 | Not stated |
A strategy tested only on the last year has never seen the 2024 market. The full providers comparison adds Kaiko and QuickNode, delivery and price.
Order-level books make limit-order fills testable
With a price-level book, a backtest assumes your limit order fills when price touches it. With every resting order in sequence, you see how much size sat ahead of you at that price and how much traded through before your turn. That turns fill assumptions into fill counts.
0xArchive rebuilds the Hyperliquid book order by order from a checkpoint and every change after it, with the wallet on each order, and replays it over WebSocket in the order it happened. The same history comes as Parquet exports for a backtest you rerun many times.
Check the window for gaps before you trust a result
0xArchive's data quality routes report completeness, gaps and cadence for each market and data type, plus known incidents with the symbols they touched. Run them on your backtest window first, and set the gaps you accept before the numbers come back.
Some limits are known and published. Hyperliquid trades before 22 March 2025 are kept but partial, and candles start on that date, so a 2023 or 2024 test should run on the book and funding instead.
Route references
Use the REST reference for order books, trades, funding and liquidations, replay and reconstruction for rebuilding a book in order, the data quality routes for gaps and incidents, and venue coverage for each market's start dates.
For current API availability and plan details, use the Hyperliquid data API, data catalog, pricing, and status.