The growth-optimal (Kelly) criterion almost surely leads to more capital in the long run and reaches levels of capital asymptotically faster than alternative strategies, but such outperformance may not be realized with high probability for an exceptionally long time. We will first demonstrate how the Kelly criterion arises in finance without first appealing to a logarithmic utility function, and then consider strategies based on alternative utilities that emphasize the probability of exceeding an underperforming benchmark faster than Kelly.

We will discuss the mathematics of portfolio optimization, assuming asset returns have a fat-tailed, alpha-stable distribution. A PDF-file of the slides is available for download.

http://www.math.uci.edu/sites/math.uci.edu/files/Fat Tailed Kelly.pdf