1 in 5000 is too small for the Riemann hypothesis, given that people have only been seriously working on it for 130 or so years, and that there still exist very smart people who are optimistic about one approach or another. (I know some of them, in fact.) It's not like P vs. NP, where the experts in the field are agreed that there's almost surely a long way to go for any approach.

I'd put it at 1% to 2.5%, myself.