Binary Ichimoku and ADX Patterns for a Beta-VAE Trading Signal

Summary

The article describes a trading signal workflow that encodes selected Ichimoku and Wilder’s ADX conditions as binary pattern flags, then trains a β-variational autoencoder to learn compact representations. Its motivation is that scaling continuous indicator values may blur discrete events such as a crossover or cloud breakout and give noisy fluctuations undue influence. A β-VAE adds a weighted regularization term to encourage a more structured latent representation. The trained model is exported for inference in an MQL5 Expert Advisor assembled with the MQL5 Wizard.

The author compares binary inputs with an earlier continuous-value pipeline and reports clearer equity behavior and a small number of accurate trades in the stated test window. The conclusion describes two years of data, with one training period and one testing period. This is limited evidence: the author calls for broader testing on real tick data and notes that the trading setup lacks proper stop management. The reported comparison therefore does not establish reliable live performance.

Key ideas

• The method converts selected Ichimoku and ADX conditions into on/off features for each bar.
• A β-VAE learns a probabilistic latent representation, with stronger KL regularization intended to encourage disentangled features.
• The trained model is exported for inference within an MQL5 Wizard Expert Advisor.
• The article reports a favorable comparison against continuous scaled inputs over its test window.
• The evidence is limited by the short evaluation, lack of proper stop management, and need for broader real tick testing.

Tags

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