Order-Flow Backtesting for High-Frequency and Hedging Strategies
Summary
The article compares conventional candlestick backtests with more detailed approaches for high-frequency and multi-instrument strategies. It explains that bars omit the timing of intrabar extremes, bid and ask quotes, and queue priority, which can distort simulated entries and hedges. It also notes that large orders may affect market prices, an effect ordinary historical replay does not capture reliably.
As an alternative, it describes a backtest based on individual trade records. Trade direction is used to estimate the best bid and ask, while simulated orders are matched against subsequent trades, with separate treatment for maker and taker orders and partial fills. The author presents results for different order sizes and sleep intervals, illustrating that simulated returns can change with order capacity and timing. The approach remains an approximation: it omits full depth and actual queue placement, simplifies immediate taker execution, and cannot fully model market impact. The stated data and simulation are limited examples, not proof of live profitability.
Key ideas
- Candlestick bars omit intrabar timing and quotes needed to model precise execution.
- Individual trade flow can help estimate best bid and ask prices and simulate order matching.
- The proposed simulator accounts for maker and taker behavior and partial fills using subsequent trades.
- Simulated returns vary with order size and strategy timing, revealing capacity effects.
- The method still simplifies queue priority, 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.