Regarding modern portfolio theory and portfolio optimization: a portfolio that allocates equal weight to all assets is often as good or better than a portfolio that results from optimization.

The statistics that are used to build a mean variance optimized portfolio are not stable. The mean is not stable and the covariance is not stable. Also, the assets are not supposed to be correlated, but they generally have a high degree of correlation.

An interesting experiment, which I have not performed, would be to build some large number of portfolios using mean variance optimization and equal weight and back test them over time. My guess is that the outcome of the experiment would not show significant advantages in return and risk between the equal weight portfolios and the mean variance portfolios.