Transaction-Level Order-Flow Backtesting for High-Frequency Strategies
Summary
This article explains why OHLC candlestick backtests can mislead when evaluating high-frequency or multi-instrument strategies. Bars omit the timing of their highs and lows, do not show the best bid and ask needed to model matching, and cannot reliably represent how a strategy’s own order size affects fills and market prices. These gaps can distort hedge timing, execution assumptions, and parameter comparisons.
The proposed alternative replays transaction-by-transaction data to infer quotes from aggressor direction and simulate maker and taker orders, partial fills, cancellations, fees, and order lifetime. An example using a crypto perpetual contract compares returns across order sizes and sleep intervals, illustrating that larger orders can receive less favorable relative outcomes under the matching model. The method is a more detailed approximation, not a complete market simulator: it still simplifies queue priority, immediate taker fills, available depth, and the market impact of replacing historical orders. The author also notes the data’s limited coverage and replay constraints.
Key ideas
- OHLC bars omit intrabar timing and quotes, limiting their value for high-frequency execution research.
- Transaction-level trade data can support more detailed estimates of bid and ask prices and order fills.
- A replay model can distinguish maker from taker execution and account for partial fills and fee differences.
- The example shows that simulated returns vary with order size and strategy timing assumptions.
- Historical trade replay still simplifies queue position, depth, immediate execution, and market impact.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.