Using minimum variance optimization with gold and spy produces a portfolio that is often 50% or more of gold. This beats the risk adjusted return of spy by a reasonable margin.

So I guess the point I'm making is that often a lot of gold makes sense too.