Define allocation and risk consistently
Let p be win probability, a the positive net win multiplier and b the positive net loss multiplier. Allocating fraction f of equity to this payoff changes wealth by 1 + f × a on a win or 1 − f × b on a loss. The unconstrained binary Kelly fraction is [p × a − (1 − p) × b] ÷ (a × b). This is the allocation fraction under those multipliers, not automatically the percentage of equity lost at a stop.
Apply the fraction and the cap
The selected Kelly multiplier scales a positive theoretical allocation before the maximum allocation cap is applied. For example, a half Kelly setting uses half the calculated fraction, subject to the cap. A nonpositive modeled expected payoff produces zero allocation. The modeled loss at the chosen allocation is f × b of equity; the calculator additionally bounds allocation below 1 ÷ b to preserve positive wealth in the loss outcome. Fees should be reflected in both net outcomes rather than subtracted from the final percentage afterward.
Test estimation uncertainty
Kelly optimization targets expected logarithmic growth in a repeated model with known probabilities. Your estimated win rate and payoff averages are not known constants. A small change in either can materially change the allocation, especially near zero edge. Check weaker probabilities, smaller wins and larger losses before interpreting the output. A cap limits the displayed allocation, but it does not account for correlated positions, exchange liquidation, sudden price gaps or an incorrect trade model. Fractional sizing reduces exposure without making an uncertain estimate accurate.