I have a Masters degree in Computational Finance and Risk Management from the University of Washington, Seattle.

Academics love mean variance portfolio optimization because the math is cool and it, supposedly, allows a quantitative portfolio construction for better investment results.

I bought into this and I have a bookshelf of books on "modern portfolio" construction.

In retrospect I am less sure about how good mean variance optimizations, even with techniques like covariance shrinkage is. We know for a fact that the assumptions are wrong. Assets are almost always correlated, since it is so difficult to find assets that are not correlated. The mean is not stable, as I mentioned.

If you run some of these tests, I very much look forward to reading about your results.