Seeing that he worked with some LTCM quants, I would imagine he knows more about the situation than you do. The kinds of ineffiencies LTCM were exploiting made sense, it's just during times of stress the historical correlations between pairs start to break.

But, hey, I'd bet on coin flips all day long if I was given favorable odds.

I don't know, the "it's just" part kind of indicates that the real probability of things remaining correlated were not what LTCM thought they were, to the point where they couldn't afford to stay in the position long enough for things to snap back. I don't buy that as "always winning".