Entropy Lab
The Entropy Lab goes beyond linear correlation to detect nonlinear dependencies between assets that Pearson completely misses.
The Problem with Pearson
Pearson correlation only captures linear relationships. Two assets can have near-zero Pearson correlation yet still be deeply dependent — just in a nonlinear way.
Example: an asset that mirrors another's volatility (not direction) shows r ≈ 0 on Pearson but high NMI.
Shannon Entropy
The Entropy module computes Shannon entropy for each asset's price return distribution:
H(X) = -Σ p(x) · log₂(p(x))Low entropy → returns are predictable, concentrated distribution
High entropy → returns are chaotic, spread distribution
Entropy Ranking
The Entropy Ranking bar chart ranks all active assets from most predictable (low entropy) to most chaotic (high entropy).
This answers: "Which asset has the most predictable price behavior right now?"
NMI Heatmap
Normalized Mutual Information (NMI) measures how much knowing one asset's returns tells you about another's — regardless of whether the relationship is linear.
Range: 0.0 (completely independent) to 1.0 (perfectly dependent)
The heatmap color scale: dark purple (0) → violet (0.5) → green (1.0)
Hidden Connections
The bottom panel highlights Hidden Connections — pairs where:
Pearson correlation is weak (< 0.3 in absolute value)
But NMI is high (> 0.4)
These are pairs that appear uncorrelated on the surface but have a meaningful nonlinear dependence. They often represent lagged relationships, volatility coupling, or regime-dependent co-movement.
Live Run Mode
Toggle Live Run to continuously update entropy calculations as new price ticks arrive. In static mode, entropy is computed once per session load.
Entropy calculations require a minimum sample size. The Assets Ready counter in the header shows how many assets have enough history to compute reliable entropy values.
Last updated