Seems to defy the odds....

[petrw1]

No factors in SOOO many attempts???

Top trial factoring producers for the last year.
This hourly report generated Mar 18 2013  4:04AM UTC
Your current rank is:        225                                  
                             Total                      |      Rank Change       | Overtake date based on 
Rank  Member Name           GHz-Days Attempts Successes |90day 30day 7 day 1 day | 90 days 30 days  7 days
-- -------------------- ------------ -------- --------- |----- ----- ----- ----- | ------- ------- -------     
37 Patrik Johansson        40582.279     5174         0 |   7    1             | >999dy   741dy   282dy     
47 Bill Staffen            28173.974     2942         6 | ***   24    5    1 |                            
63 Jaxon                   16948.251     1778         1 |  16    5    2       |  317dy   301dy   306dy     
65 Tim S                   15892.279     1521         1 | 644    8    4       |                            
79 Sebastian Hornbostel     9747.261    10450         0 |  22          2       |                           
216 ANONYMOUS               1281.248      670         0 |  36   19    9       |          120dy   118dy

[LaurV]

- bad hardware
- wrong doing (mfaktX fake reports, for raising the credit)
- or pure and simple bad luck...

edit: separate reports? I knew for example Xyzzy used to report only the factors in the past, because of the difficulty - at that time - of reporting lots of "no factor" lines. How about if someone thinks "viceversa" and reports only "no factor", having a separate account where he reports "factors"? I can't see a reason for it, beside of boasting to friends "look how many factors I found", because I can not imagine that one would intentionally want not to report factors... Secondly of finding a 100 million digit prime (an bringing in the EFF money), you help GIMPS mostly when you find a factor (well, arguably :D) and reduce the burden of doing two LL tests... (saving power, green planet, avoid global warming bla bla.. )

[dbaugh]

And this guy is just really lucky.

Rank Member Name GHz-Days Attempts Successes
13 ANONYMOUS 114646.974 405808 240344

[LaurV]

Originally Posted by dbaugh

And this guy is just really lucky.

No, that is very normal, you take some 800M or 900M from 65 to 66 and find a lot of factors. Many "old participants", who started with the beginning of gimps, have done "low bitlevel" (like from 40 bits to 65 bits) for the ranges we now do LL and DC, and they got a success every few exponents. Well, approximately. Other people did "low B1 P-1", with third party tools, and there is still no way to report a "NF-PM1" outside of P95, so they only reported the factors.

Anyhow, high success rate is not "smelly". You can't cheat PrimeNet by reporting more factors than you can find. Low success rate is smelly, and dangerous, because fake reports result in skipping bitlevels which can have factors for some exponents, resulting in more LL and DC for other people, and for the GIMPS in general.

[Mini-Geek]

Originally Posted by petrw1

No factors in SOOO many attempts???

                             Total                      |      Rank Change       | Overtake date based on 
Rank  Member Name           GHz-Days Attempts Successes |90day 30day 7 day 1 day | 90 days 30 days  7 days
-- -------------------- ------------ -------- --------- |----- ----- ----- ----- | ------- ------- -------     
65 Tim S                   15892.279     1521         1 | 644    8    4       |

Hey, that's me! After looking into it, it looks like the factors I'm finding with mfaktc through GPU to 72, which are reported as manual testing, are being mislabeled as F-PM1 results. This might explain the others listed as well, especially if they are paired with unusually high P-1 success percentages, such as my ~8.3%.

GPU to 72 reports that from 15.329 expected factors, I've found 16, so I'm right on track with slightly-above-average luck and a slim chance of bad hardware.

Originally Posted by LaurV

Anyhow, high success rate is not "smelly". You can't cheat PrimeNet by reporting more factors than you can find.

It is "smelly" in the sense that the person might not be sharing all of their failures with PrimeNet, thus causing duplicate work to be done.

[Dubslow]

Is this not the same patrik who posts the error rate plots? I wouldn't be surprised if it was some specialized work he's doing. I rather doubt it's anything malicious.

[kracker]

 Rank  Member Name              GHz-Days Attempts Successes |90day 30day 7 day 1 day
    40 Patrik Johansson         2338.475      695       120 |   3    1    1

In P-1.