Backtesting with Binance L2 data: what to check before a strategy run

By Cruscy Data · 2026-10-30

Binance order-book backtesting starts with an execution question: what could your order have done against the recorded liquidity? A candle-only experiment cannot answer that question with the same detail. Before interpreting a return, write down the data window and the assumptions used to turn decisions into fills.

Start with a reproducible window

Record the exchange, market, symbol, UTC start and end, stream names and coverage. Use data that was available at each decision time. A completed candle's high, low and close become known after that candle closes; making them available earlier introduces information the strategy would not have had.

Cruscy Data exposes stream schemas and dates on the dataset pages. Its current backtest workflow supports spot strategies. Access to futures market-data streams should not be confused with a claim that the spot runner simulates perpetual margin, funding or liquidation rules.

State the costs and fill model

Include the commissions and execution assumptions actually used by the run. Distinguish a recorded quote from a simulated fill. For a market order, depth and available size matter. For a resting limit order, reaching a price does not necessarily establish your queue position or guarantee execution.

Historical L2 increases the information available to a simulator; it does not make every execution assumption exact. If latency, queue priority or hidden liquidity is approximated, label the approximation in the report. A result that depends on an untested fill assumption needs further investigation.

Inspect defects before comparing strategies

Look for missing book updates, delayed ingestion and incomplete intervals. Do not silently fill gaps with later observations. Review the published quality reports and preserve a record of any excluded interval. Spot and futures quality checks can have different coverage; use the report for the market being studied.

Compare more than the final balance

Keep the trade count, fees, drawdown, position exposure and the benchmark alongside net return. Reserve a later period for checking a strategy that was developed on an earlier one. Repeatedly choosing the best result on the same window changes what that window can tell you.

A practical first run is a small strategy on the documented free spot demo hour. Follow the strategy contract and API examples, inspect the resulting report, and expand the window once you understand the records and the runner. A successful technical run is a working experiment; its trading outcome still has to be measured.