How to Diagnose Recursive Indicator Variance in Freqtrade

Summary

Recursive analysis helps check whether indicator values depend materially on how many startup candles are available. Recursive formulas use prior values, so an indicator calculated over the full backtest history may differ from one calculated in a dry or live run with a shorter exchange-provided history. The page uses a cumulative counter example to show why the latest value can shift when the calculation window changes.

Freqtrade’s command first calculates benchmark indicator values over a selected historical range, then repeats calculations using specified startup candle counts and compares the final-row values. The output reports percentage variance by indicator; insufficient history may produce an unavailable value. The documentation recommends ample benchmark history and explains exchange candle limits that can constrain startup counts. The tool does not backtest a strategy, inspect the effect on entries or exits, or assess indicators calculated outside the documented indicator-population stages. Its lookahead check is limited to indicator values, so a full lookahead analysis is still needed. Small variance should be judged against the strategy’s decision thresholds; exact zero is not always necessary.

Key ideas

• Recursive formulas can make indicator values sensitive to the amount of startup history used.
• The command compares final-row indicator values from multiple startup candle counts with a benchmark calculation.
• The output helps identify whether a longer startup history may be needed, but users must judge the practical importance of variance.
• The analysis requires historical market data, and exchange API limits can constrain how many candles are fetched.
• The command does not backtest trades and does not provide a full lookahead-bias assessment.

Tags

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