Rule

Treat any Kelly number on a high-win-rate, negatively-skewed payoff as an overestimate.

What it means

Kelly has a structural preference for high win rates, which always arrive with negative skew - exactly the shape of a credit spread. Full Kelly carries a one-in-three chance of halving the bankroll before doubling it.

Related, and more useful

Positive skew VALIDATES FASTER. At a per-trade Sharpe of 0.25 a positively-skewed structure needs about 28 trades to prove itself; a negatively-skewed one needs 64. For an agent that must earn evidence before it is allowed size, long-convexity structures are cheaper to learn from - a statistical argument, independent of the risk one.