New report: Cost per Factor Found...

[TheJudger]

Originally Posted by diamonddave

Can we have clearer data?

*hmm* not really.

But here are some stats from ~7900 P-1 runs I've done (M50.000.000-M61.000.000) with proper B1/B2 values:

• 42 factors found with size 2⁶⁹ - 2⁷⁰
• 49 factors found with size 2⁷⁰ - 2⁷¹
• 35 factors found with size 2⁷¹ - 2⁷²

for ~7900 exponents you expect 7900/69 = 114 factors for TF from 2⁶⁹ to 2⁷⁰. Now let us remove those 42 factors found via P-1: 114 - 42 = 72. So we've a 72/7900 = 1/110 chance for finding a factor using TF from 2⁶⁹ to 2⁷⁰.

Numbers for 2⁶⁹ - 2⁷⁰: 7900 / 69 - 42 = 72 ==> 72/7900 = 1/110
Numbers for 2⁷⁰ - 2⁷¹: 7900 / 70 - 49 = 64 ==> 64/7900 = 1/161
Numbers for 2⁷¹ - 2⁷²: 7900 / 71 - 35 = 76 ==> 76/7900 = 1/104

I know that some (I guess < 10%) of my 7900 P-1 assignments had TF to higher than usual TF limit so those numbers are even more complicated.

7900 assignments are at least one magnitude to small for some real good stats.

Oliver

[diamonddave]

Originally Posted by TheJudger

*hmm* not really.

But here are some stats from ~7900 P-1 runs I've done (M50.000.000-M61.000.000) with proper B1/B2 values:

• 42 factors found with size 2⁶⁹ - 2⁷⁰
• 49 factors found with size 2⁷⁰ - 2⁷¹
• 35 factors found with size 2⁷¹ - 2⁷²

for ~7900 exponents you expect 7900/69 = 114 factors for TF from 2⁶⁹ to 2⁷⁰. Now let us remove those 42 factors found via P-1: 114 - 42 = 72. So we've a 72/7900 = 1/110 chance for finding a factor using TF from 2⁶⁹ to 2⁷⁰.

Numbers for 2⁶⁹ - 2⁷⁰: 7900 / 69 - 42 = 72 ==> 72/7900 = 1/110
Numbers for 2⁷⁰ - 2⁷¹: 7900 / 70 - 49 = 64 ==> 64/7900 = 1/161
Numbers for 2⁷¹ - 2⁷²: 7900 / 71 - 35 = 76 ==> 76/7900 = 1/104

I know that some (I guess < 10%) of my 7900 P-1 assignments had TF to higher than usual TF limit so those numbers are even more complicated.

7900 assignments are at least one magnitude to small for some real good stats.

Oliver

Thanks allot, it wasn't obvious to me at first sight that P-1 had such an impact.

[petrw1]

I think I now have a large enough sample size to be statistically realiable....300 P-1 completions.....

Most of my P-1s start with the chance of finding a factor somewhere around 4-4.5%. But after 300 completions I have 8 factors found= 2.67%.

The project as a whole is right at 4.5%: 60 out of 1330...guess it's just me

[Dubslow]

It's not just you.

I've done 8 P-1 and found 2 factors. Very ridiculous.

In return, I've done 135 exponents 3 bits each and only found one factor. As I understand it I would expect about 6 (~=135/71*3).

But that is the nature of statistics.