User brilong
 

I just looked around in the [URL="http://www.mersenne.org/report_recent_results/"]stats[/URL] a bit and in the trial factoring top producers list I found this user brilong to have finished 629 assignments woth over 10000 GHzd with none of them having a factor.
That seems very odd to me, and even calculating an average of 1/65 Chance of finding a factor per assignment it shoult have a Chance of 0.005% of finding no factors for 629 tests - maybe he is bringing in these results just to push his statistics? I think somebody should have a look at that.

edit: Matthew P. Del Buono looks similar and Sebastian Hornbostel too, the first two of these seem to have joined just recently.

Sometimes PrimeNet misunderstands the source of some factors. If you look at the [URL="http://www.mersenne.org/report_top_500_p-1/"]top P-1 list[/URL], you'll see that brilong has 12 successes out of 12 attempts. In all likelihood, these are actually TF factors. 12 out of 642 TF attempts is perfectly reasonable (1 in 53.5, which should be better than average).

I was bored so, for some reason, I checked the estimation of 1/65 =/
(Looking at the first 500 values of (Successes/Attempts) in the [URL="http://www.mersenne.org/report_top_500_tf/"]top TF list[/URL], there's a mean of 0.016259481158623995 with stdev 0.044598467127841386, and median of 0.014438339986704773.
When ignoring those with less than 10 successes, there's a mean of almost 0.019, and when ignoring those with less than 20 successes the mean is almost 0.02 and median is about 0.0655. Needless to say, all of the information above is uninteresting.)

Well, in the very worst case, he is in fact bullshitting his results to get more credit, and the result is we would have to do an LL test on an exponent which would have given a factor.

While unfortunate, this wouldn't be the end of the world. We're close enough to the optimal TF level that it's small potatoes.

[QUOTE=TheMawn;371288]Well, in the very worst case, he is in fact bullshitting his results to get more credit, and the result is we would have to do an LL test on an exponent which would have given a factor.[/QUOTE]

Likely the opposite -- Primenet is (still) mis-interpreting factors found by GPU TF'ing as being done by PM-1 (unless a "No factor found" result is turned in first), and thus he's actually being under-credited.

Primenet does not assign for LL'ing or DC'ing candidates which have been factored.

[QUOTE=manfred4;371271]edit: Matthew P. Del Buono looks similar and Sebastian Hornbostel too, the first two of these seem to have joined just recently.[/QUOTE]

LOL... A last minute edit by a participant. An age old trick by a certain David.

With regards to [URL="https://www.gpu72.com/reports/worker/2d6d30cc0b735a709e316193b0f7fb78/"]Matthew P. Del Buono[/URL], looks pretty nominal to me.

With regards to [URL="https://www.gpu72.com/reports/worker/7fe2e567d20835388ee8c0428e4efd04/"]Sebastion Hornbostel[/URL], again, looks pretty nominal.

Any other questions?

Okay, I only looked at the one site, I didn't recognize that the factors may have been reported but not picked up by PrimeNet properly. Thanks for the info.

[QUOTE=TheMawn;371288]Well, in the very worst case, he is in fact bullshitting his results to get more credit, and the result is we would have to do an LL test on an exponent which would have given a factor.[/QUOTE]

[QUOTE=chalsall;371290]Likely the opposite -- Primenet is (still) mis-interpreting factors found by GPU TF'ing as being done by PM-1 (unless a "No factor found" result is turned in first), and thus he's actually being under-credited.

Primenet does not assign for LL'ing or DC'ing candidates which have been factored.[/QUOTE]

I believe we didn't completely understand each other. If he was in fact submitting a bunch of no-factors results, the only effect of that is we have to do a few LL tests that we really wouldn't have needed.

As far as the project progress is concerned, bullshit no-factors results aren't really a big deal.