Track 11Closed
Signal Denoising (Wavelet + Kalman)
denoising price (wavelet/Kalman) improves the directional entry
Key metric
standalone 0/8 validation (Kalman velocity is useful as a tool)
- Why an edge should exist / hypothesis
- Less noise in the signal → cleaner direction. The classic DSP approach.
- Construction
- Wavelet denoising + Kalman velocity as a direction picker.
- Test data
- BTC, 8-slice validation.
- Results
- Wavelet denoising is harmful (non-causal look-ahead). Kalman velocity is the only useful component: direction accuracy 54% vs RSI 47%, PF 1.25 vs 0.73. But as a standalone strategy — 0/8.
- Validation
- Standalone 0/8; Kalman kept as a tool in other branches.
- Root cause
- Wavelet pulls in the future (non-causal). The direction gain is real but insufficient for a standalone edge.
- Verdict
- Dead as a strategy; Kalman velocity preserved as a reusable tool.
- Source
archive/SIGNAL_DENOISING_RESEARCH.md
Want the same check on your strategy?
Validation →