Diagnosing Empty Training Data in Simulated Strategy Runs

Summary

The discussion addresses a platform issue where a strategy can complete a one-day backtest but fails in simulated trading because its training data is empty. The proposed diagnosis focuses on how the M3 module extracts data and whether its date range is bound to a trading day. Binding the training data to a single trading day may leave too little history available in simulation, according to the respondent’s hypothesis.

The response recommends leaving the M3 training data unbound while binding the prediction data. It also cautions that the explanation is based on experience rather than a confirmed experiment, so the suggested configuration needs testing in the user’s environment. The exchange supplies no code changes, reproduced error trace, or verified outcome. Its value is a specific troubleshooting hypothesis about separating training and prediction date settings, rather than a confirmed general rule for all strategies or platform setups.

Key ideas

• A strategy may backtest successfully yet encounter empty training data in simulated trading.
• The respondent suspects that binding M3 training extraction to a trading day can narrow its available history.
• The proposed setup leaves training data unbound and binds the prediction data.
• The explanation is explicitly unverified and should be tested in the relevant platform configuration.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.