This is survivorship bias. He is not wrong in that yes, we should consider ideas from domain experts carefully and not be overly dismissive. But it is absolutely survivorship bias to believe that every crazy idea from a domain expert worked out.
1) That if the crazy idea is from a domain expert, it deserves to be taken seriously. It doesn't deserve to be taken as gospel, but it shouldn't be dismissed outright.
2) He's betting (but not stating as fact) that if you did blindly support the crazy ideas of domain experts, you'd be right more often than you're wrong.
This is not the same thing as saying domain experts are always right. Simply that all things being equal, you're better off betting on their expertise over your own ignorance. But even better than that is to look into the matter until you have sufficient expertise to properly evaluate their idea.
> Such ideas are not guaranteed to work. But they don't have to be. They just have to be sufficiently good bets — to have sufficiently high expected value. And I think on average they do.
It's likely that you're both right. If something is biased 45% off of a coin toss, but gives 25x returns, it's still positive EV.
This is a part of VC and speculative investment that a lot of people gloss over. Success is the difference between being wrong 95% of the time and being wrong 97% of the time. (And, with Softbank around, sometimes it's the difference between being wrong 97% of the time and 99.5% of the time.)
Where the author suggest a rule: "Such ideas are not guaranteed to work. But they don't have to be. They just have to be sufficiently good bets — to have sufficiently high expected value." - it is hedged against survivorship bias.
Who believes that?