The app for independent voices
The portfolio framework that contains all other portfolio frameworks.
GARP (Generalized Adaptive Rank Portfolio) just dropped, and its central claim is bold: HRP, ERC, IVP, 1/N, HERC, and BL-HRP are all special cases of a single recursive equation.
Think of it as the GARCH of portfolio construction: it doesn't compete with existing models… it contains them.
What makes it different:
No distributional assumptions: operates entirely on ranks, so Gaussian priors, heavy tails, and skewness are all irrelevant to the math
No hard cluster cuts: weights accumulate continuously top-down through the full dendrogram; no arbitrary k
Plug-and-play signals: swap momentum for alpha, swap VIX for a macro signal; the architecture doesn't change
Crisis-adaptive by design: a single scalar p(t) ∈ [0,1] continuously shifts the distance metric from correlation to co-crash (lower tail dependence) and tilts allocation away from contagious assets
Preliminary precursor results (not yet full GARP): Sharpe of 1.887, max drawdown of -7.13% over 684 weeks out-of-sample, with a 75% drawdown reduction during the 2020 COVID crash relative to equal weighting.
Full empirical evaluation is in a companion paper. This one is pure theory… and the theory is clean.
