major stats weirdness

[Moo_the_cow]

From the "Project stats" page -
Remaining tests n > 4000000 214386 tests

This seems awfully low, considering that the # of tests remaining after n>4M
was over 500000 just a week ago. I know that sieving removed some of them, but i'm baffled at how any method can reduce the number of remaining candidates by half.



:confused:

[jjjjL]

my bad. :blush:

fixed.

-Louie

[Keroberts1]

there have been a very large # of test listed as completed today? Hows this hapeneing? It doesn't appear that the # of test below 4 million has gone down by that much in the last 24 hours is the srtats completeld test indicator counting secret/supersecret too?

[Nuri]

Originally posted by Keroberts1
is the srtats completeld test indicator counting secret/supersecret too?
Yes.

[eatmadustch]

just a minor thing that I don't think looks too good (while we're talking about stats!)
the "maximum size of next prime" is 6020602 digits. I guess this is 31337? How about not counting 31337 for the stats (or at least largest primes?). BTW, isn't 6 million rather large? I thought you were "just" trying to get the record, but thats somewhere near 3.5 million! Or is this an error?

[Keroberts1]

stats counting secret and supersecret seems pointless they are not the same sized tests as the meaningful proths and i don' think they should be in the stats updates it just misleads people and it looks like were doing alot more work than we are. Its also making it difficult to accuratly access how fast the project is moving along

[Keroberts1]

Does anyone else have any input on this matter?

[Frodo42]

Not really any new input other than that i agree with you keroberts. I myself do a little work on secret (when I have my windows partition running) and would not mind not being able to see the stats for secret, I get reports about the progress every time my computer get a new test (which is quite often).
Secret and supersecret corrupts the "number of tests completed" stats (they easily complete a couple of hundreds a day) so if possible I would prefer to have them removed from the stats, or at least from the "number of test completed" stat.

[Mystwalker]

I have a system running as supersecret. Although it seems like I'm the only one running that account ATM, it affects the stats somewhat. I don't think that it's useful to include those special accounts into the stats.

btw. Louie, is it possible to delete the pendings of the secret and supersecret accounts more often (maybe every 2 days for secret ans every hour for supersecret)? E.g. a cron job should to be job, shouldn't it?
That would make the extended stats table more readable - assuming those accounts stay in the main stats area...

[Keroberts1]

While doing a little playing around i found a couple more problems with the stats. at about 535 today the stars run updated and i noticed that whil the number of tests completed today had only risen by 7 the number of total test completed rose by 14. This is very weird. With all of the misdirecting information in the stats it is very difficult (if not completely impossible) to figure out how fast we are moving along. Is this a known error, is there somthing being done about it? If anyone else would like to comment on this please do.

[Keroberts1]

I may have found my mistake wit hthe stats reading. The last day meant last 24 hours and perhaps 7 of the test had been finished at a time that was just 24 hours before so they were removed fro mthe tally still, I nbelieve the stats do need a little work.

[Keroberts1]

a possible reply to eatmadustch's question about the current size of the largest prime. that value corresponds to a prime that would be from an n value about 19999990 or ~ 20000000. My guess is that Louie has either expanded this ot be the largest prime the search could ever find or he's started assigning the top values ot the 31337 account so he can not just beat the current largest prime but so he can blow it away. This is just a guess if Louie would like to explain more we'd all be happy ot hear from him I'm sure.

[sobo]

I dont know if this is a coincidence or not, but the max size of a prime that could be found is pretty close to the numbers that GIMPS is currently testing.

[Nuri]

As Keroberts1 suggested, the 6020602 digits test is one of the largest numbers in our current sieve range of n=3m to 20m. It seems to me that, it should be something around 19,999,750 or so.

It seems like eight of us tried the 31337 (http://www.seventeenorbust.com/stats/users/user.mhtml?userID=6) account at the beginning, but nobody continued, with the last block on 17th. If that's the case, and if nobody else starts to run that account within next few days, then Maximum size of next prime stat will return back to normal (1.1 million +), on 27th.