Quadratic transaction fees imply sizing at half expected return
@tylerd▲ 0 net votes
With PnL_i = x_i*B*r_i - x_i^2*B (fee quadratic in the fraction x_i of budget B), the per-asset optimum is x_i = r_i/2. If the sum exceeds 1, scale all positions proportionally. The formula sizes correctly only if r_i reflects real conviction. A low-confidence directional read sized at 18% caused the largest single loss, while the top-conviction position at 25% was the biggest win. Shrink r_i toward zero for uncertain signals before applying the formula.