The brain produces a probability for every symbol (e.g. "NVDA: 67% likely to go up in the next hour"). Knowing the probability isn't enough — you need to know how much to bet.

The Kelly Criterion is the math that maximizes long-term compound growth. Full Kelly = (probability × win-amount − loss-prob × loss-amount) / win-amount. The catch: full Kelly assumes your edge estimate is perfect. If your estimate is even slightly off, full Kelly can blow up your account.

This page uses confidence-scaled quarter-Kelly: we take the pure Kelly fraction, multiply by 0.25 (safety margin), then multiply by additional confidence factors:

• Uncertainty — wider MC dropout interval = smaller position
• Cross-method agreement — if multi-horizon + bootstrap + k-NN disagree, smaller position
• Conformal halfwidth — wide rigorous-coverage intervals = smaller position
• BSS trust — if the brain isn't learning (BSS < 0), tiny positions
• Source reliability — if Stooq is degraded, smaller positions

For multiple positions: when 3 A-tier signals fire at once, naive sizing would over-allocate because of correlation. The portfolio allocator estimates correlation between symbols (NVDA and AMD are 80% correlated — they move together) and scales each position so total portfolio risk stays under your cap.

For options (not just stocks): see Options 101 for Beginners for how to translate "buy 50 shares of NVDA" into "buy 1 NVDA call at the right strike."

Bottom line: a high-probability prediction + tight stops + conservative sizing + compounding wins = a portfolio that grows over time WITHOUT a single bad trade wiping you out.