Primenet ECM-Fermat results

ET_ · 2018-08-21, 11:23

I remember that, for a while, the server automated the request for ECM-F workload giving units without the "factors known" tail, thus causing the "rediscover" of known Fermat factors. That issue has been corrected on the actual Primenet server.

Another issue is that you can manually report the same result line many times: this has two side effects:

1 - the actual number of curves actually done on each Fermat number grows without the curves having been done.
2 - someone on the Primenet server keeps cheating, increasing their position on the list just submitting over and over the same lines.

Now, consider the following actual listing (365 days list):

Code:

                                      Total                     | 
 Rank  Team Name                    GHz-Days Attempts Successes |
------ ------------------------ ------------ -------- --------- |
     1 ComputerraRU               208494.846    42657       301 |  (Apparently all curves done by Yxinity)
     2 GIMPS.Russia                29344.945     5567         0 |
     3 Porta Kruse                 19197.372      260         0 |
     4 Ozzie Prime                  8036.409      544       132 |  (all curves done by David Bessel)
     5 zeit                         7780.461      141         0 |
     6 GIMPS.Occitania              7651.522      109        23 |  (all curves done by Jacques MOLNE)
     7 Team_Italia                  7215.657      109         0 |
     8 UKRAINE                      6050.696     5437      5625 |  (all curves done by Apsen)
     9 PrimeNet                     2199.626      114        15 | 
    10 More Power!!!                1971.083       61         0 |
    11 Team-Deutschland             1919.614       37         0 |
    12 One Man Team                 1823.617       25         0 |
    13 GIMPS.California              911.755       23         0 |
    14 Hungary                       895.216       13         0 |
    15 Ars Technica Team Prime Rib   792.136       39         1 |
    16 GIMPS Team Philippines        787.904      698       867 | (all curves done by LazyDude)
    17 Cap'n Entropy's Cavalcade...  773.498        9         0 |
    18 Czech Vainlich Team           671.638       10         0 |
    19 Team_Bammann                  613.656        7         0 |
    20 SHADO                         608.254       16         0 |
    21 Wile E. Coyote                551.924       40         7 |
    22 GIMPSChina                    475.401       42        14 |
    23 X Grubbers Kick Ass           282.107        7         0 |
    24 His Dark Materials            162.623        2         0 |
    25 sannerud.com                  157.447      116       150 |  (all curves done by sannerud.com)

Someone has more "successes" than attempts, i.e. they seem to have found more factors than curves done.

I am not claiming that the listing has been obtained by forging the results, I'm just hinting for the moderators' attention on such incongruence, and asking if such technical behaviour could be amended in any way (first thing that comes to my mind: insert a "sigma" column to avoid reporting the same curve over and over).

Luigi

henryzz · 2018-08-21, 11:41

My guess is that returning a composite factor could produce multiple successes.

LaurV · 2018-08-28, 03:14

Nothing unusual, some people just do batch work without reservation, and they only report back when they found factors. You can't "cheat" much, that way, a factor is a factor is a factor. I would be more concerned (like 0.137% instead of 0%, hehe) about the guys with thousands of attempts and no success, they smell fishy to me. (Do not forget that it is still very easy to falsify ECM "no factor" results, and some "credit whores" may just want to see their name in top100 or so).

On the other hand, if you talk about known Fermat factors (ECM-F), their number is quite limited, and it is extremely #oddlookingtome*, to see in the top100 list more "successes" than the number of known factors... unless of course, some people found factors and they keep them secret, hehe.

------
* trying to keep the pace with the young generation, hashtag, blah blah... although we don't really know what all those new things mean...

GP2 · 2018-08-28, 12:40

There are only 344 known Fermat factors. F12 has six prime factors, and some people keep finding the smallest five and their composite combinations over and over because they didn't include the known-factor list in their worktodo line. I think the latest versions of mprime automatically supply these factors but not everyone uses a recent version.

2018-08-21, 11:23	#1
ET_ Banned "Luigi" Aug 2002 Team Italia 2⁶×3×5² Posts	Primenet ECM-Fermat results I remember that, for a while, the server automated the request for ECM-F workload giving units without the "factors known" tail, thus causing the "rediscover" of known Fermat factors. That issue has been corrected on the actual Primenet server. Another issue is that you can manually report the same result line many times: this has two side effects: 1 - the actual number of curves actually done on each Fermat number grows without the curves having been done. 2 - someone on the Primenet server keeps cheating, increasing their position on the list just submitting over and over the same lines. Now, consider the following actual listing (365 days list): Code: Total \| Rank Team Name GHz-Days Attempts Successes \| ------ ------------------------ ------------ -------- --------- \| 1 ComputerraRU 208494.846 42657 301 \| (Apparently all curves done by Yxinity) 2 GIMPS.Russia 29344.945 5567 0 \| 3 Porta Kruse 19197.372 260 0 \| 4 Ozzie Prime 8036.409 544 132 \| (all curves done by David Bessel) 5 zeit 7780.461 141 0 \| 6 GIMPS.Occitania 7651.522 109 23 \| (all curves done by Jacques MOLNE) 7 Team_Italia 7215.657 109 0 \| 8 UKRAINE 6050.696 5437 5625 \| (all curves done by Apsen) 9 PrimeNet 2199.626 114 15 \| 10 More Power!!! 1971.083 61 0 \| 11 Team-Deutschland 1919.614 37 0 \| 12 One Man Team 1823.617 25 0 \| 13 GIMPS.California 911.755 23 0 \| 14 Hungary 895.216 13 0 \| 15 Ars Technica Team Prime Rib 792.136 39 1 \| 16 GIMPS Team Philippines 787.904 698 867 \| (all curves done by LazyDude) 17 Cap'n Entropy's Cavalcade... 773.498 9 0 \| 18 Czech Vainlich Team 671.638 10 0 \| 19 Team_Bammann 613.656 7 0 \| 20 SHADO 608.254 16 0 \| 21 Wile E. Coyote 551.924 40 7 \| 22 GIMPSChina 475.401 42 14 \| 23 X Grubbers Kick Ass 282.107 7 0 \| 24 His Dark Materials 162.623 2 0 \| 25 sannerud.com 157.447 116 150 \| (all curves done by sannerud.com) Someone has more "successes" than attempts, i.e. they seem to have found more factors than curves done. I am not claiming that the listing has been obtained by forging the results, I'm just hinting for the moderators' attention on such incongruence, and asking if such technical behaviour could be amended in any way (first thing that comes to my mind: insert a "sigma" column to avoid reporting the same curve over and over). Luigi Last fiddled with by ET_ on 2018-08-21 at 11:25

2018-08-28, 03:14	#3
LaurV Romulan Interpreter Jun 2011 Thailand 2438₁₆ Posts	Nothing unusual, some people just do batch work without reservation, and they only report back when they found factors. You can't "cheat" much, that way, a factor is a factor is a factor. I would be more concerned (like 0.137% instead of 0%, hehe) about the guys with thousands of attempts and no success, they smell fishy to me. (Do not forget that it is still very easy to falsify ECM "no factor" results, and some "credit whores" may just want to see their name in top100 or so). On the other hand, if you talk about known Fermat factors (ECM-F), their number is quite limited, and it is extremely #oddlookingtome, to see in the top100 list more "successes" than the number of known factors... unless of course, some people found factors and they keep them secret, hehe. ------ trying to keep the pace with the young generation, hashtag, blah blah... although we don't really know what all those new things mean... Last fiddled with by LaurV on 2018-08-28 at 03:22

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Primenet Recent Results Request	garo	PrimeNet	7	2014-02-03 02:53
When did PrimeNet begin rubberstamping results?	Chuck	PrimeNet	64	2014-01-06 01:22
making PrimeNet accept duplicate TF results?	ixfd64	PrimeNet	5	2012-12-08 07:04
V5 not reporting results to PrimeNet	edron1011	Software	8	2009-01-26 16:59
Problem with Reporting Results to Primenet	RMAC9.5	Software	3	2003-11-02 23:06

2018-08-21, 11:41	#2
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 2·2,909 Posts	My guess is that returning a composite factor could produce multiple successes.

2018-08-28, 12:40	#4
GP2 Sep 2003 A17₁₆ Posts	There are only 344 known Fermat factors. F12 has six prime factors, and some people keep finding the smallest five and their composite combinations over and over because they didn't include the known-factor list in their worktodo line. I think the latest versions of mprime automatically supply these factors but not everyone uses a recent version.