Six-Gate Filtering for LSTM Trading Signals

Summary

The article describes Jardine's Gate, a sequential filter intended to decide whether an LSTM-generated cryptocurrency CFD signal should be traded. The six checks cover market structure through compression-based entropy, agreement among multiple LSTM experts, signal confidence, regime-adjusted probability, alignment with a 200-period EMA, and a loss-streak kill switch. A signal must clear every gate to reach execution, with cheaper checks placed earlier to reduce computation.

The author reports that 127 of 43,200 monthly signals passed the filter and that those trades had a 100% backtest win rate. These figures are claims from the article, not independently validated evidence; the supplied text is incomplete in its discussion of the later gates and does not provide enough detail to assess testing methodology, sample robustness, or live performance. The thresholds are presented as tuned for BTCUSD on a one-minute chart, limiting how readily they transfer to other markets or settings. The filter is a system design example, not proof that independent gates eliminate losing trades.

Key ideas

• The six-gate pipeline filters signals by market structure, expert agreement, confidence, adjusted probability, trend alignment, and loss streak.
• Entropy is estimated from compression behavior and used to block signals in low-structure conditions.
• The design requires every gate to pass and orders inexpensive checks early in the sequence.
• The article reports a perfect backtest win rate for a small subset of a much larger signal stream, but gives limited validation detail.
• Thresholds are described for BTCUSD on a one-minute timeframe and may not generalize to other settings.

Tags

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