View Single Post
Old 2019-12-24, 16:56   #6
mart_r
 
mart_r's Avatar
 
Dec 2008
you know...around...

67710 Posts
Default

Been re-reading "Prime number races" by Granville and Martin and "Cramér vs. Cramér" by Pintz again, also some of Maier's work. (The more I read it, the better I understand it.)
Conclusion: Might as well go with
M\hspace{1}=\hspace{1}Ri(p'+\frac{q}{4})-Ri(p+\frac{q}{4})

It's close enough to what I previously thought was the most accurate way of measuring the merit and quite easy to calculate. A trade-off, so to speak.

q=188940 / p=8356739 / g=76*q qualifies as CSG>1 even by g/log²(p')/\varphi(q)

If anyone's interested, I'll post a more exhaustive list of gaps with CSG>1 when measured by the above formula.
mart_r is offline   Reply With Quote