Checking Recursive Indicator Sensitivity to Startup Candle Count

Summary

Recursive indicators depend on preceding values, so their outputs can change when calculations begin with different amounts of historical data. A backtest may use a long requested date range, while a live or dry run may have fewer candles available from the exchange. The page illustrates how this mismatch can alter an indicator’s latest value and potentially make historical and live results diverge.

Freqtrade’s recursive-analysis process calculates indicators with a benchmark history and with several startup candle counts, then compares their values on the final candle. The resulting percentage differences help users judge whether the chosen startup history is long enough for each indicator. The command needs downloaded historical data, does not backtest trades, and only evaluates specified indicator calculations; it does not establish whether observed differences change entries or exits. Its simple lookahead check also covers indicator values only, so a separate full lookahead analysis is needed. Users must choose acceptable variance in the context of their strategy rather than assume zero variance is necessary.

Key ideas

• Recursive indicators can produce different values when initialized with different amounts of historical data.
• Different history lengths in backtests and live runs can cause indicator values and results to diverge.
• Recursive analysis compares final-candle indicator values across startup candle counts against a benchmark calculation.
• The output reports indicator variance, which users must assess for its potential effect on trading decisions.
• The command does not backtest trades and does not replace a full lookahead-bias analysis.

Tags

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