But you shouldn't also pretend you have zero information. They hit the barge last time, engineering principles are sound, so those two things alone give you an estimate of "quite likely". Yes, 80% is kind of arbitrary, but it tells you the same story as "quite likely" with an added benefit of being able to plug it into some math that will yield you better results than going with just words.
(Note that I think that starting with a good 50/50 coinflip and revising it based on evidence in this case is reasonable, I just think it's nice to have a reasoned opposing opinion sometimes.)
>But you shouldn't also pretend you have zero information.
If you've never done something before you really have no idea if it's going to work, and in rocket launch everything that occurs after the last failure is a complete unknown in terms of failure modes.
Now, if he'd made successful flyback stages before, or it was a relatively routine thing for the industry he might have a good enough feel to assign a ballpark number. But that's not the situation.
In the link you provided he uses computer failure as an example, where a person who is familiar with computers does have some information - I've owned my current rig for three years and it's never failed. If you told me you thought it has a 50% chance of failure next month I can pretty confidently say that's an overestimate.
But it's useless to pull numbers out of the air for something as complicated as SpaceX is trying to do.
If you have an expectation, you have a probability. Not a numeric one, but a probability nonetheless.
Nobody has ever landed a rocket before. But we don't expect the rocket to turn into an alarm clock or suddenly develop antigravity. In fact, we have quite reasonable expectations on what behaviors the rocket will exhibit. We might, for instance, expect the rocket to crash more than we expect it to land. What else does that say rather than p(crash) > p(land)?
(Of course, this may well come down to "probability as ratio" vs. "probability as anticipation", which is probably a matter of preference.)
See "If It’s Worth Doing, It’s Worth Doing With Made-Up Statistics": http://slatestarcodex.com/2013/05/02/if-its-worth-doing-its-....