Strategy research

Walk-forward: weights are refit on prior data before each test fold using the live plan's exact pipeline, then simulated net of estimated pool slippage against buy & hold, equal-weight, and all-Root. Short crypto histories make every Sharpe noisy — the ± band shows one standard error.

The plan above is a deterministic max-Sharpe optimization. Click Explain this rebalance for an AI brief — the headline Sharpe change, the concrete buys/sells, and what the math is blind to.

Max-Sharpe over (holdings ∪ candidates ∪ Root): μ = shrunk carry + shrunk drift − liquidity penalty, risk = constant-correlation shrinkage covariance, sleeve sized to a vol target with caps. Click Show for the full method, or expand a row for its μ/σ breakdown.