Regime Lab

A single Sharpe ratio hides the truth that most edges are conditional: great in calm, trending markets and dead in turbulence (or the reverse). The Regime Lab breaks a strategy's history into volatility regimes and shows how it did in each — so you know when your edge actually works.

Open cBots → Regime Lab (/quant/regimes).

What it does

Given a return series (or equity curve, oldest first), it:

• computes a trailing realized volatility at each point and splits the history into Calm, Normal and Turbulent regimes by the terciles of that volatility;
• reports per-regime performance — observations, mean return, volatility and Sharpe — so you can see where the edge lives;
• estimates the Hurst exponent via rescaled-range (R/S) analysis: above ~0.55 the series is trending / persistent, below ~0.45 it is mean-reverting, and around 0.5 it is close to a random walk.

POST /api/quant/regimes
{ "returns": [...], "window": 10 }   // or { "equity": [...] }

Why it is reliable

Pure, deterministic domain code (Core.Regimes) with no infrastructure dependency and no external calls — unit-tested for regime separation (calm vs turbulent volatility) and for the Hurst direction (anti-persistent series score below 0.5, a persistent trend scores above). The same regime signal feeds the autonomous agents' reflection loop, so an agent can lean into the regimes where its edge is real.