"the portfolio that maximizes an adjusted CAGR" over the ranges 2015-2025? Isn't that one just massively overfitting to the unique geopolitical events of 2016-current? 1970-2025 sounds better. But still that's baking in the USD exchange rate.

What happens if you backtest it by finding the equivalent portfolios for 1998-2008 (or even 1919-1929?)