View Single Post
Old 2013-10-17, 12:44   #22
TheCount's Avatar
Sep 2013
Perth, Au.

11000102 Posts

Well, I completed checking the probability formulas for the 1k's. What I found is that:
- henryzz new method is the same as the old. Same inputs and results.
- these methods require the range to be pre-sieved. Only 21% of the 1k's are sieved.
- my new method don't require pre-sieving and is 98.9% correlated on the Riesel side and 97.0% correlated on the Sierpinski side.
- I don't know which method is more accurate.

The weight (w) I used in the spreadsheet column O, attached, is:
w = k*b*P1e6 / (li(110,000*k*b)-li(100,001*k*b))

Where li(x) ∼ x/ln(x) + x/ln(x)² + 2x/ln(x)³ + 6x/ln(x)⁴ + 24x/ln(x)⁵ + ...

CRUS - 1k'

In the early 1970s, following a counting session, the Count would laugh maniacally, "AH AH AH AH AH!"
TheCount is offline   Reply With Quote