PCA Reconstruction Error
Principal Component Analysis reconstruction error — measures how well current market conditions fit historical patterns. High error = regime change.
Why AlgoZilla Uses PCA Reconstruction Error
Anomaly detection: when PCA error spikes, something unusual is happening. The model becomes more cautious and requires higher conviction for entries.
Feature Variants
AlgoZilla expands every base indicator into multiple variants: raw values, delta (rate of change over 8/12/24 bars), divergences, and multi-timeframe computations across 2H, 4H, 8H, 12H, and 24H horizons. This is what sets AlgoZilla apart: 170+ features, each retrained every two weeks per coin.
Variants: 1 variant (raw value)
Part of a Bigger Picture
No single indicator drives AlgoZilla decisions. This is one of 170+ features feeding into a machine learning ensemble, retrained every two weeks per coin. The model learns which features matter in each market regime.
