Again Silver is using data, he believes in the 'party decides' theory that the best early indicators are endorsements (from currently elected politicians) and Trump has zero. This cycle will test the 'party decides' theory. Silver is also right that early polling, especially early national polling has not been a good indicator. However we are now transitioning into the time when it isn't early polling and Trump still has a lead in NH and is in second in IA, so history is slowly but surely moving in Trump's direction to at least be a serious contender.
Considering Bush heavily leads among party insiders from his initial "heir apparant
status, and his nearest competitor in that regard is Rubio, i think that theory is already quite debunked even if Trump loses.