Kelly fraction estimation for multiple correlated bets

It is well-known that expected portfolio growth is maximized by maximizing expected logarithmic utility. This investment criterion is known as Kelly betting. It has many optimality properties but is considered to be risky. Blackjack teams and other advantage gamblers practice a fraction of the Kelly optimal to decrease risk. Some hedge fund managers are thought to practice according to Kelly principles. We use a continuous multivariate Geometric Brownian motion model and present an interval estimate for the historical fraction for a portfolio of correlated bets, possibly including a risk-free asset. Historical data comes from a range of sources and the results provide a risk aversion statistic, which corresponds to an isoelastic utility function and level of relative risk aversion