Why Live Trading Can Diverge from Backtests and How to Handle It
Summary
This guide explains operational differences between backtests and simulated or live brokerage trading that can cause strategies to behave differently. It highlights initial account holdings, delayed or unfilled orders, missing holding-period fields, manual trades that alter strategy positions, cancellations, and delayed execution or reports. The examples show how code that assumes immediate fills or complete internal position records can fail when actual account state differs.
Suggested mitigations include reconciling initial holdings, avoiding untracked external trades, using safer dictionary lookups, maintaining holding-period records, and accounting for pending orders and cancellation states. The author also describes platform improvements such as strategy-level asset separation and order or fill events, but these are presented as planned changes at the time of writing. The examples are platform-specific and illustrative rather than measured evidence; they do not quantify execution costs or establish that any particular fix suits every strategy. The central lesson is to model asynchronous execution and account state explicitly when moving from historical simulation to live trading.
Key ideas
- Backtests may assume initial holdings are absent, orders fill immediately, and reports arrive without delay.
- Live accounts can contain positions and trades that the strategy did not create.
- Brokerage data may omit holding-period fields available in backtests.
- Strategies should track pending orders, fills, cancellations, and external position changes explicitly.
- The examples explain failure modes but provide no quantified performance comparison.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.