Sieving success as a heuristic for odds of a prime
 

I'm busy sieving 32767 * 2 ^ n - 1 very deeply now (and busy testing it at the first few millions of bits or so - each time narrowing down the boundaries of the sieve domain).

Well deeply -that's as far as the hardware allows here. Nowhere nearby what PG manages to achieve of course.

What i notice is that to my surprise it keeps finding factors and removing exponents also at higher bitlevels.

Whereas other riesels with a similar nash weight dry up much sooner.

Would the guess be logical to assume then odds for a prime is also much lower than similar nashed k's?