Comment by elliptic

Out of curiosity, how much better, typically, is the 'optimal' blending rule than blending with some random or 'reasonable' weights? Never really thought to ask myself that question...