Commodities triangulation sandbox.

How the triangulation works. The engine self-calibrates from the loaded five-year history: drift comes from a momentum vs mean-reversion blend (lag-1 autocorrelation decides how much the recent 26 weeks dominate), volatility comes from an EWMA estimator (lambda 0.94) that captures volatility clustering the way GARCH-class models do, and a regime detector places today's 12-week realized volatility inside its own five-year distribution. Node A projects the damped blended trend. Node B runs 400 geometric Brownian motion paths on the clustered volatility, plotting the median and the p10 to p90 band. Node C builds a stepwise vector from your liquidity bias, jumping at simulated policy events. The triangulated point is the weight-blended endpoint of the three vectors, and the structural boundary guard clamps it into your scenario envelope. Because the envelope is an input, the output always reflects your assumption; treat it as a way to explore "what would the path look like if", never as a prediction.