Page 1 of 5 12345 LastLast
Results 1 to 40 of 172

Thread: sieve stat question

  1. #1

    sieve stat question

    if a factor is found and a test is currently issued for that N value the nwill the test be aborted? If not then will the test be reassigned if the person who is assigned the proth test doesn't finish it or will it then be aborted, Finally (thge point I'm trying otget at) if the test is deleted after someone lets it expire then will the sieve stats show the factor as having saved a test even though the test was still assigned.

  2. #2
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    if a factor is found and a test is currently issued for that N value the nwill the test be aborted?
    There are some other threads discussing this issue, but the answer is no - currently the SB client has no way of being signalled to abandon a test.
    If not then will the test be reassigned if the person who is assigned the proth test doesn't finish it or will it then be aborted
    I am 90% sure it won't be reassigned, but Louie would need to say for sure.
    Finally (thge point I'm trying otget at) if the test is deleted after someone lets it expire then will the sieve stats show the factor as having saved a test even though the test was still assigned
    I've been thinking about this one recently. Assuming the next.txt file doesn't drop sudenly (as has happened once), then as things stand right now the factor won't score until it enters the double check window (at which point it will receive full score). Clearly this isn't fair, because the factor was useful and did save a test. But with this situation it is difficult to determine when to say for sure that a PRP test has been saved. I am planning to add another magic point which will equate to the 'n upper bound' - if a factor is below this point and has 0 PRP, then it will score full value. This still means it could take >6 months before the factor scores, but that will be a lot better than waiting for it to enter a double check window (which could be never!).

    Mike.

  3. #3
    thanks not what i was hopping ot hear but still its nice ot be able ot get feedback on questions like this

  4. #4
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Just been trying to tidy up the user references in the sieve stats.

    I've combined Kroberts5 and Keroberts1, which will now be displayed as Kroberts5 (Keroberts1). As things were, anything where you were logged-in went to Kroberts5, anything where you're not logged-in went to Keroberts1 due to the association with the co-ordination thread. If this is wrong, let me know, and it can be undone easily, or done differently.

    Next question ceselb and smh, are you the same person? If so, would you like me to combine, the two, and which one as primary?

    Mike.

    P.S. Tully, have you tried the lottery lately? Of your first two factors sieved, one is straight on target, and if you'd submitted two days later in would have been in the PRP zone!
    Last edited by MikeH; 08-18-2003 at 03:50 PM.

  5. #5
    thanks you're great

  6. #6
    Moderator ceselb's Avatar
    Join Date
    Jun 2002
    Location
    Linkoping, Sweden
    Posts
    224
    No, I'm not smh. Where did you get that idea?


  7. #7
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    No, I'm not smh. Where did you get that idea?
    P-1 co-ordination thread:

    4151500 4152000 smh/ceselb 23 0.505595 3 [completed]


  8. #8
    Moderator ceselb's Avatar
    Join Date
    Jun 2002
    Location
    Linkoping, Sweden
    Posts
    224
    Ahh. smh couldn't do it all, so I took over the last part. Just wanted to get him some credit for it.

    Sorry about the confusion.

  9. #9
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Interesting stats comment.

    There is still one k value for which no one has managed to find a factor which scores >3500.

    k=22699: mklasson 4847188223273203 | 22699*2^353350 (scored 3500.00000)

    When will this one fall?

  10. #10
    Junior Member
    Join Date
    Feb 2003
    Location
    Linköping, Sweden
    Posts
    17
    Originally posted by MikeH
    Interesting stats comment.

    There is still one k value for which no one has managed to find a factor which scores >3500.

    k=22699: mklasson 4847188223273203 | 22699*2^353350 (scored 3500.00000)

    When will this one fall?
    *cough*
    k=22699: Xrillo 54968301363529 | 22699*2^4720582 (scored 3827.83695)

  11. #11
    Senior Member Frodo42's Avatar
    Join Date
    Nov 2002
    Location
    Jutland, Denmark
    Posts
    299
    Sieve stats are currently unavaileble
    You have reached this page through an incomplete address for the site you are looking for. To search the server for the site, please enter the address in the field below and click Submit.
    I hope it will be up running soon again, i just found 4 factors using P-1

  12. #12
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Sieve stats are currently unavaileble
    To quote my ISP
    Some PIPEX customers are reporting that their websites are not displaying correctly.

    Customers trying to view their website on one of the following URLs:
    www.username.dial.pipex.com
    www.username.dsl.pipex.com
    are seeing a 'welcome to webhosting' holding page instead of their usual website contents.

    We are currently investigating this issue with our network engineers.

    PIPEX apologises for any inconvenience caused by this problem.
    Sorry, but at least they claim to be investigating. Upload is working fine, so when it does come back online it'll be up to date.

  13. #13
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    I hope it will be up running soon again, i just found 4 factors using P-1
    Frodo, your P-1 efforts sure are showing up the sievers right now. You have found 9 on target factors in under a week. My luck seems to have gone down hill - my last good factor was Tuesday, since then, with 100+ factors submitted, my best is 30 points for a double check factor.

  14. #14
    Sieve it, baby!
    Join Date
    Nov 2002
    Location
    Potsdam, Germany
    Posts
    959
    Well, can't check the site either, but have something that could probably fixed.
    The active window at the scores.htm page has a working (read: updating) lower bound, but the upper bound stays fixed. As one of my factors is going to get into the active window soon, I'd like to know if it's just a display error.

  15. #15
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    The active window at the scores.htm page has a working (read: updating) lower bound, but the upper bound stays fixed.
    It's OK, all part of my closing the window down from 500K to 200K - which will then level the playing field in terms of big scores for sieve and P-1. Because I didn't want the top of the window to jump, it is holding at 4830000. It will start to change when the bottom of the window reaches 4630000. And given the PRP effort is moving at about 10K/day, that should be in about 18 days time.

    In the end we'll all still score the same, just that for sieving most factors will be swept in the window rather than landing straight in. But this does mean that if P-1ers follow the instruction of selecting a range 200K above next n, then they have a good chance of featuring on the max scores tables.

    Mike.

    P.S. I can see the scores right now, not sure how long it'll last.

  16. #16
    Senior Member
    Join Date
    Feb 2003
    Location
    Sweden
    Posts
    158
    MikeH,
    is there something wrong with the "Factors next to pass 'n upper bound'" stats? Pretty much no factors at all seem to go there. I have none listed, for example, while I certainly have submitted lots of factors between the n upper bound and the active window. Or am I just misunderstanding what it's for?

    EDIT: *whistling* As usual I should probably have spent a minute more thinking before posting. =) I take it the only factors listed there are those that haven't had any PRP test done, i.e. those that were found for a k whose upper bound was slightly lower than the maximal upper bound of any k?

    Mikael

  17. #17
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    I take it the only factors listed there are those that haven't had any PRP test done, i.e. those that were found for a k whose upper bound was slightly lower than the maximal upper bound of any k?
    Correct. And this is the slightly depressing list, because so far all those that have appeared and then disapeared have done so because PRP tests have been returned. We have yet to have one make it down to the 'n upper bound', although I think we are reasonably confident that Xrillo's and Xeltrix's should make it.

  18. #18
    Sieve it, baby!
    Join Date
    Nov 2002
    Location
    Potsdam, Germany
    Posts
    959
    Could it be that the second active window uses the 'supersecret' stats instead of 'secret'? The first one stays fixed at 400,000...

    btw. there are still some character creation errors, like 0M&ltn< 1M
    I guess there's only a ';' missing.

    But that are just cosmetic things - when it comes to functionality (and motivation ), the scoring page is incredible!

  19. #19
    Senior Member
    Join Date
    Feb 2003
    Location
    Sweden
    Posts
    158
    Louie,

    any plans on having the prp client abort the test when submitting a block for a test that has a known factor? Seems like it should be an easy fix considering that the client aborts tests wrongfully at times (the "server had no record of test" thing).

  20. #20
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Could it be that the second active window uses the 'supersecret' stats instead of 'secret'? The first one stays fixed at 400,000...
    Again correct, but different comment to Mikael's. Since Louie placed cap of 400K on the DC PRP, currently we have paused at 400K. Not sure what Louie's plans are for DC. I remember reading a sugestion some time back that DC should be kept at 1/10 of main PRP effort - that sound like a very good idea to me. It also means that DC will always take 1% of main PRP effort which is insignificant but useful.

    btw. there are still some character creation errors, like 0M<n< 1M
    I thought I'd fixed all the issues with these. Can you let me know exactly which line on which page, and I'll take another look.

    Mike.

  21. #21
    Sieve it, baby!
    Join Date
    Nov 2002
    Location
    Potsdam, Germany
    Posts
    959
    That's right, DC isn't concerned right now, but a factor found for a k/n-pair soon to be checked by the secret account should still be worth some reward, as it lowers some effort, doesn't it?

    Problems with character conversion:

    - User main page: Largest Score Unique by n
    - individual page: titles of Incomplete and Complete (Plus, there's a typo in "Resevred ranges" - hell yeah, I am german )
    - project page: Remaining candidates and unique factors by n (n&lt20M)

    Hm, I don't know what browser you're using, but maybe that one is "intelligent" enough to just correct it.
    Well, gonna slap some mosquitos now and get some sleep. G'night!

  22. #22
    I love 67607
    Join Date
    Dec 2002
    Location
    Istanbul
    Posts
    752
    Mike,

    I dunno how hard it would be to implement these ideas (and also how useful some of them would be ), but would you consider adding some more features like;

    - adding a column that shows daily change in score (and average daily change over last 14 days) at user and team main pages

    - adding the breakdown of new unique factors on an individual and team basis, and again, (average daily change over last 14 days) at user and team main pages

    - a breakdown of daily score change in the individual pages. (i.e. score change from yesterday's submissions and score change from previous submissions). If I understood the scoring system correctly, at the time being the effect of previous submissions is limited to the factors that are submitted above the 90% sieve point. However, things will start to change slowly when main active window upper bound starts to move upwards from the 4830000 within next two weeks. This is also true for the DC active window, though the impact would be less. Score from previous submissions will start to become more and more significant within the total change in individual scores, and I think it would be interesting to see the dynamics from this perspective as well.

  23. #23
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    That's right, DC isn't concerned right now, but a factor found for a k/n-pair soon to be checked by the secret account should still be worth some reward, as it lowers some effort, doesn't it?
    Yeah, and these do score - in fact they receive full score because they are marked as PRP=0. If you get one you'll notice, because they'll score >50 points. I've had a few of these, indeed you can tell when I made the change so that these scored by looking at my changed scores

    Thanks for the pointers, I might get chance later today to take a look, problem is every time a add something I forget the HTML rules (IE6.0 BTW).

    Nuri, thanks for the suggestions, all sound like good ideas to me, I'll have a think.

  24. #24
    Sieve it, baby!
    Join Date
    Nov 2002
    Location
    Potsdam, Germany
    Posts
    959
    Yeah, and these do score - in fact they receive full score because they are marked as PRP=0
    Ok, now I understood you point. Strange I didn't see it before...

  25. #25
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Mystwalker, give the stats pages a try now. Acording to this HTML checker all is now good (but I haven't looked at the gap pages yet).

    There were lots of errors.

  26. #26
    Sieve it, baby!
    Join Date
    Nov 2002
    Location
    Potsdam, Germany
    Posts
    959
    Almost all gone:

    - Project page: remaining candidates and unique factors by n (n&lt20M) <-- title + table head + left column
    - gap page: top navigation (currently looking like this when using Opera: <&ltprev next>> &ltalt n range>)

  27. #27
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Along the line of Nuri's request (but only a tiny part), here is a score card for all users, except that it's a lot bigger - 100 of each category and 500 recent factors. So now if you're wondering where those 400+ factors found today have come from, now you know...

  28. #28
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    - adding a column that shows daily change in score (and average daily change over last 14 days) at user and team main pages
    Again, not quite what you were asking for, but on the main score page there's a "User scores (daily average for last 14 days)".

    Very interesting numbers.

  29. #29
    Senior Member
    Join Date
    Feb 2003
    Location
    Sweden
    Posts
    158
    MikeH,
    very nice! I seem a bit unlucky though... Good thing the active window starts moving again soon. *rubbing hands*

  30. #30
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    very nice! I seem a bit unlucky though
    I think the only thing that's kept my score up is indeed a bit of luck and a small dose of P-1.

  31. #31
    Moderator ceselb's Avatar
    Join Date
    Jun 2002
    Location
    Linkoping, Sweden
    Posts
    224
    Speaking of scores, this seems a bit off imho. Will it really score that high, or is it a glitch?
    Code:
        p (T)      k         n      Score     Factor found    Score changed  Score was    Score could be  Reqd bias
     9280.677   4847   4830063     35.000  Wed 10-Sep-2003                                    676605.103     232.02

  32. #32
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Speaking of scores, this seems a bit off imho. Will it really score that high, or is it a glitch?
    It's no glitch. If the bias can get to 232, then it will score this much. Of course the bias will never reach this (sieving 90% point would need to be at 9280T, when currently we're at 71T)

    So in reality this one's score will be determined by the bias on the day before it drops into the PRP window.

  33. #33
    Moderator ceselb's Avatar
    Join Date
    Jun 2002
    Location
    Linkoping, Sweden
    Posts
    224
    Originally posted by MikeH
    If the bias can get to 232
    ...
    Of course the bias will never reach this
    ...
    So in reality this one's score will be determined by the bias on the day before it drops into the PRP window.
    Whouldn't it be better just to pick a slightly (or even a lot) saner bias? Or am I missing something here?

  34. #34
    Sieve it, baby!
    Join Date
    Nov 2002
    Location
    Potsdam, Germany
    Posts
    959
    The bias is a lot lower for smaller factors. The bias takes into account that it's getting harder to find factors in higher ranges for sieving. Plus, it takes care that factors found through factoring don't score more than a just sieved factor would.

  35. #35
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Whouldn't it be better just to pick a slightly (or even a lot) saner bias?
    The bias requred for any given factor to achieve it's maximum score is p/40T, so in your case 9280/40 = 232.

  36. #36
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Nuri said
    - a breakdown of daily score change in the individual pages. (i.e. score change from yesterday's submissions and score change from previous submissions).
    Done. Next I'll have a think about those main tables and beefing up the team stuff.

  37. #37
    We now have a stat saying what tests have had a factor foundbut are still being performed at this moment. Could we add a list of the users that have bee nassigned that test so someone can let the user know that their test should be aborted?

  38. #38
    Junior Member
    Join Date
    Feb 2003
    Location
    Linköping, Sweden
    Posts
    17
    I recall having found a factor for an ongoing test a while ago. Kinda' wondering what happend to it. It is no longer showing on the list of such factors/tests, and I recall someone saying in a thread that he/she aborted (but didn't manually expire it) the test after seeing the factor on that list.

    Basically, what I'm saying is should I have expected a score increase when the test expired automatically?

    Maybe the test was reassigned to someone and completed? This seems unlikely though.

    -- Added:
    I did some research and found this by MikeH:

    "The other pieces of good news are we now get a chance to confirm that this test won't be re-assigned, and assuming it doesn't, Xrillo you'll get full score when the 'n upper bound' catches up!"

    The 'n upper bound' hasn't caught up with it yet. Someone who knows his way around the stats might know something about this.

    Last edited by Xrillo; 09-16-2003 at 06:51 AM.

  39. #39
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Maybe the test was reassigned to someone and completed? This seems unlikely though.
    I took a look at this a few days ago when I noticed it had disappeared from the table. In the results.txt file, the factor in question is now show as having 1 PRP result returned.

    I'm hoping that maybe the PRP=1 is not true, and is somehow a minor bug in the assigment code. But my worst fear is that this there is a much more serious bug in the assignment code, and this test really was reassigned even though a factor was present at the time of reassignment.

    Only Louie or Dave can answer this one.

    (k=5359, n=4387702)

  40. #40
    Senior Member
    Join Date
    Jan 2003
    Location
    UK
    Posts
    479
    Today I found and submitted a very strange factor

    69.077 33661 2975400 1911.054 Thu 18-Sep-2003

    Note the score - it's big considering this should only be valid for the DC window some time in the distant future. If you take a look in the results.txt, it's there with PRP=0. How can this be possible?

    According to the project stats the n upper bound has been around the 4M mark for a while. So how can the same database claim that no test has ever been performed for a candidate at 2.9M? I guess the only possibility is that a factor already exists for this candidate, but isn't present in the results or lowresults files.

    If this really is the first factor for this candidate, then how many other candidates are there that have not been tested?



    Edit: And before anyone asks, no this isn't a secret test. For this k, secret tests should have topped out at n=645336.

Page 1 of 5 12345 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •