EDAT or entropy driven adaptive trading model (needs rewiring doesnt currently work)

calculates ε (micro-noise): variance of 1-min |Δmid| or tick-imbalance volatility (upticks–downticks over 30–60s). ζ (turbulence): realized vol or Parkinson (HL), plus HL-entropy (histogram entropy of HL ranges). φ (info energy): volume × |Δprice|. ω (cycle): short Welch PSD peak ratio or 10–40 bar Hilbert phase coherence. η (adaptive weight): 1 / (1 + rolling σ_returns). Aux: order-book pressure (Σ depth_ask – Σ depth_bid at top 5), CVD slope.

Latent state (ξ) HMM with 3–4 states trained on [ε, ζ, φ, ω, book-pressure].

τ (state duration) Run-length since last state change → feeds confidence and position cap.