The issue with Kelly is that it gives you the EV (expected value) not the median outcome. So, if you are going to use Kelly position sizing .5X Kelly is much closer to optimal as it gives you a very high percentage of the expected outcome, with far less variance in terrible outcomes. What Kelly really shows is that you need to create a portfolio of uncorrelated bets that have 0 or negative correlations to one another. Kelly shows that betting everything on equities or equities and bonds is an incorrect approach to optimizing outcomes. For investing Sharpe ratio is useful as it gives a more directly translatable use of something very similar to Kelly. For example, if you model out Monte Carlo simulations for a .5 Sharpe 50% target vol is optimal, at 1 Sharpe 100% target vol is optimal, at 2 Sharpe 200% target vol. Of course, similar to Kelly you would need to halve that to eliminate the wild negative outcomes.