Thanks for the post! What are your thoughts on the limitations of mean-variance optimization techniques, a cornerstone of MPT, when looking at asset classes that have a non-normal return distribution? Is variance, or standard deviation, still then the most applicable measure of "risk" when history has shown us time and again that risky asset return distribution have significant negative skewness and excess kurtosis? One alternative may be mean-VaR optimization, but I've yet to see its implementation discussed by practitioners involved in the portfolio construction process.