Bloom Filters and RNNs for Noise-Aware Trailing Stops in MQL5

Summary

The article outlines a custom MQL5 trailing module that uses a Bloom filter to screen repeated prices and a recurrent neural network (RNN) state to moderate stop adjustments. Price values are converted to point-based integers, hashed into a compact bit array, and treated as duplicates when the relevant bits are already set. This is intended to reserve computation for price changes deemed novel and reduce stop movement in response to small fluctuations.

The RNN maintains state across bars or ticks and applies a threshold to influence stop decisions. The article describes the class structure and initialization, but the supplied text omits much of the RNN implementation and test details. It reports that a moderated test increased gross profitability while also increasing equity drawdown, underscoring a tradeoff and the need to tune the threshold to risk preferences. Bloom filters can produce false positives, and point-level price deduplication may be sensitive to instrument precision and filter settings. The approach is presented as a prototype for testing, not as evidence of reliable live performance.

Key ideas

• A Bloom filter can screen previously observed price values using compact bit storage and hashed membership checks.
• The example rounds prices to symbol points before hashing, so the chosen price granularity affects which moves count as new.
• An RNN hidden state carries information between updates and moderates when the trailing logic responds.
• The reported moderated test paired higher gross profitability with greater equity drawdown.
• The article presents a prototype and leaves threshold calibration and live performance uncertain.

Tags

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