Compounded growth versus bet size is always a hump, never a line: betting 0% of the bankroll earns nothing, betting everything repeatedly guarantees the ruinous loss eventually arrives, so the curve must rise, peak and fall. The peak is full Kelly — gamblers call the same idea optimal-f.
The interactive bet is shaped like premium selling: it wins 85% of the time but the rare loss is 5x the win. Full Kelly lands near just 17% of the bankroll, and around 2x Kelly the long-run growth rate crosses zero — betting twice the optimum turns a positive-edge bet into a losing strategy through size alone.
Real traders run fractional Kelly because edges are estimated rather than known, unmodeled fat tails pull the true peak left of the computed one, and full-Kelly drawdowns (routinely ~50%) are deeper than most humans can hold. Half Kelly keeps roughly three-quarters of the growth for about half the swing.
In an options account, Kelly arrives through mundane dials: define the bet's loss as the crash-scenario assignment damage per contract, choose the fraction of the account that tail can't kill, and let contract count be the rounding — down, even to zero. Buying power is never the fraction; wealth is.
Educational, not investment advice.