Nightly build-off · entry B

The Correlation Tax

Markowitz says flat staking across correlated bet legs is mathematically dominated. Drag the A↔B correlation slider and watch the "diversified" flat point fall off the efficient frontier, live.

arXiv:2306.12639 — Efficient Solution of Portfolio Optimization via Dimension Reduction & Sparsification

Bet legs

Edge = expected return on stake. Vol = standard deviation of outcome. A & B are same-game props (QB passing TDs ↔ his TE's receiving yards) — naturally correlated.

Correlation: Leg A ↔ Leg B (the same-game pair)0.35

This is the lever. Crank it up and watch the flat-stake marker drop inside the frontier while the optimizer starves Leg B.

Full correlation matrix

Every other pairwise correlation — also live, also feeding the solve. Diagonal is fixed at 1.

Covariance matrix is PSD — no correction needed.

Aggressiveness (κ)4.0

Which point on the frontier is "yours" right now — 0 = safest (global min-variance mix), max = all-in on the single best edge.

Efficient frontier expected return vs. volatility, recomputed every tick

frontier flat stake your κ portfolio max-Sharpe

Stake allocation flat (1/N) vs. optimal at your κ

Flat-stake volatility

—

at the flat portfolio's own return

Excess risk vs. frontier

—

same return, risk you're giving away for free

Leg B weight: flat → optimal

—

the correlated leg gets starved as ρ rises

Sharpe: flat vs. your κ

—

return per unit of volatility

DIVERSIFIED

Loading…