Page 1 of 5 12345 LastLast
Results 1 to 40 of 180

Thread: Input wanted on planned new algoritm from users perspective

  1. #1

    Input wanted on planned new algoritm from users perspective

    The new more complex algorithm is coming along now, and I wanted to get some input so we can get any concerns out in the open now while its still being built. As currently planned, here are the major changes on the user's side that you would experience:

    - resumes where it left off when you quit - it will only upload after a batch of 5000 thus and not everytime you start/quit
    - upload will still be approx. same size as it is now
    - you can quit any time immediately
    - first batch will be 5000 (you have no control, but see above)
    - following batches will be smaller (maybe 200) but take longer to build
    - henceforth we refer to a 'batch' as a generation
    - more points will be awarded for later generation structures to encourage people to get there rather than sticking at generation 0 which is faster to generate - roughly take the sqrt of the gen # to get the scale factor for the points
    - each generation will build from the best structure of the previous generation (i.e. the method is iterative)
    - each CPU/instance will run its own set of independent generations etc. and you 'best RMSD' will be the best from any generation on any CPU/instance, but the latest generation best structure from each CPU/instance will be stored by us
    - when generation 50 is reached, it will restart at zero again
    - server can make cool 'folding movies' from the 50 generations which you can then download and watch
    - as a side effect, each machine will need to be uniquely identified so it will be not be easy to, say, generate on one machine and then upload from another - this is necessary to avoid other nasty potential problems but should not affect people you use proxy servers or firewalls, only physically moving around data will cause trouble.
    - you will at most be able to buffer 50 generations (about 2 days work (maybe)) but this could change

    Ill add more stuff as I think of it. Comments and complaints are welcome. Keep in mind some concessions will be necessary to get this new more complicated algorithm to work, which may include alienating some users with special needs.

    EDIT:

    Actually, I think we can do it without those last two points, so scratch those off
    Last edited by Brian the Fist; 01-15-2003 at 06:06 PM.
    Howard Feldman

  2. #2
    Senior Member
    Join Date
    Jul 2002
    Location
    Kodiak, Alaska
    Posts
    432
    I can see a few things that might need working out.
    At work, the first system to download and start running the new client first gets stopped, send all the tiny number of folds back, and then the directory is copied to C:\Folding2. I then stop the other machines, and copy this to their drive, kill c:\folding and rename this to c:\folding.
    (If I ever start running large numbers of systems again, I know I can try and get the proxy server to work - but for 2-3 machines I have on for the whole protein - that equal the 10 I was running, I'll refrain from trying to figure out what was wrong with the proxy server setup.)
    When I get new systems to test out - (Brand new Dells or slightly older 256Meg+ machines) I'll copy this same directory on new systems and run it overnight (or 3 weeks, if they take forever to pick it up).
    ----
    So it would be nice if the program doesn't get hard coded for the machine it was installed on.
    ----
    At least one member of my team is using a computer at his wive's work that isn't connected to the internet. Once a week, he'll stop by to get the filelist & .bz2 files, (delete and reinstall the program) take them home, copy them into a blank folding directory - and run the copy of DF in that directory to upload the folds.
    Others have mentioned using "sneakernet" approaches on a weekly basis.
    ----
    Will there be a way to upload sneakernetted folds, and over a longer period than 2 days for those machines not connected to the internet? (or will we have to start bringing them back seed information to get them to start up again after the first "generation 50" sequence is finished?)

  3. #3
    Junior Member
    Join Date
    Apr 2002
    Location
    The Knights Who Say Ni!
    Posts
    27
    " - as a side effect, each machine will need to be uniquely identified so it will be not be easy to, say, generate on one machine and then upload from another - this is necessary to avoid other nasty potential problems but should not affect people you use proxy servers or firewalls, only physically moving around data will cause trouble.
    - you will at most be able to buffer 50 generations (about 2 days work (maybe)) but this could change"

    This could drastically reduce my output.

    The first item wouldn't be a huge deal AS LONG AS one can copy the whole DF folder from one machine (say a non-internet enabled server) to an internet enabled PC and successfully upload the results. Make sense? But, if one can only upload from the machine that produced the results, then I am screwed.

    The second item (2 days / 50 generations max before needing to upload) will hurt. If I need to harvest -nonetted results every two days, then I will need to find a new project. I don't have time for that.

    Upwards of 10Ghz (and more at times) is only available to me in a -nonet type setup.

    rg

  4. #4
    Fixer of Broken Things FoBoT's Avatar
    Join Date
    Dec 2001
    Location
    Holden MO
    Posts
    2,137
    i am not a complainer

    what ever is best for the project is fine

    with that said, if the work can't be moved (easily or hard) from the original machine to another to send, it may impact my ability to participate

    but i can be very creative and am not afraid of obsticles , so again, whatever is best to get the job done (see sig )

    my situation really lends itself to projects that allow/accomidate sneakernetting :/


    more later
    Use the right tool for the right job!

  5. #5
    Member vsemaska's Avatar
    Join Date
    Mar 2002
    Location
    Connecticut, USA
    Posts
    49

    Re: Input wanted on planned new algoritm from users perspective

    Originally posted by Brian the Fist
    - as a side effect, each machine will need to be uniquely identified so it will be not be easy to, say, generate on one machine and then upload from another - this is necessary to avoid other nasty potential problems but should not affect people you use proxy servers or firewalls, only physically moving around data will cause trouble.
    - you will at most be able to buffer 50 generations (about 2 days work (maybe)) but this could change

    Ill add more stuff as I think of it. Comments and complaints are welcome. Keep in mind some concessions will be necessary to get this new more complicated algorithm to work, which may include alienating some users with special needs.

    EDIT:

    Actually, I think we can do it without those last two points, so scratch those off
    According to the 'EDIT:' line it sounds like Howard decided that unique IDs and 50 generations limit won't be needed.

    I too 'sneakernet' numerous systems so Howard I think you should keep this in mind since it sounds like a lot of people here do this.

    Vic

  6. #6
    TeAm AnandTech Insidious's Avatar
    Join Date
    Aug 2002
    Location
    Indianapolis, IN.
    Posts
    96
    If I have understood this thread correctly, sneakernetting WILL be possible. and you AREN'T planning on limiting us to 2 days worth of work between connections. Is this right?... If so, I don't see much difficulty with the changes.

    I think the loss of participants would be dramatic without the above.

    I also am wondering how the new scoring system will correlate to the scoring system that is in place at present. Is this new system going to make it impossible to close large gaps in scores created when the clients were fast? (ie: will newcommers have no chance of ever catching up?) That might cause a bit of heartburn for up & commers and/or discourage new participants from joining up.

    Finally, I have been watching some 'new-commer' projects commit suicide with premature releases of systems that are unstable and/or disfunctional causing loss of work and/or inability to get new work as needed.

    I am begging you..... please don't let this happen to DF!!!!!!!

    -Sid
    ~~~~ Just Passin' Through ~~~~

  7. #7
    Member vsemaska's Avatar
    Join Date
    Mar 2002
    Location
    Connecticut, USA
    Posts
    49
    Howard,

    Since this new algorithm is such a dramatic change have you considered doing a beta release to a small no. of participants. Let everyone else keep running with the current software until the new one seems stable. I'd be willing to test the new software on the platform I use.

    Vic

  8. #8
    Originally posted by vsemaska
    Howard,

    Since this new algorithm is such a dramatic change have you considered doing a beta release to a small no. of participants. Let everyone else keep running with the current software until the new one seems stable. I'd be willing to test the new software on the platform I use.

    Vic
    Actually, that is exactly what we intend to do. Although based on the dismal turnout for the screensaver 'beta test', i don't know how well it will work.

    Anyhow yes, we should be able to get the sneakernet working fine.

    Also I forgot to mention, stats on Team pages will also list total for current protein and overall so newcomers will still have something to show off (but top 10 will still be overall - I might add a top 10 for current protein as well though, we'll see).

    The scoring will be as fair as possible and have thought of and will implement ways to discourage people from trying to, for example, repeatedly generate and upload generation 49 structures (which are worth more). You could keep generating generation 0 structures (which are faster to generate that all others) but this is why you'll get less points for them - maybe even none - they are just the 'stepping stone' required to get into generation one where the real work begins. Generation 0 structures are like what you're making right now.

    Actually I like that. How would people feel if you got zero points for gen. 0 and then got scaled points for future generations as mentioned above? It takes only about 1/2hr-2 hrs. to make the first 5000 for most people, and then days to make the rest, so you wouldn't get credit for that couple hours work, but it would be required to reach gen. 1 where you DO start getting credit. I think this is fair and will encourage people to get up high in the generations. Ok, where are you stats-ho's? Does this sound reasonable?
    Howard Feldman

  9. #9
    Junior Member
    Join Date
    Apr 2002
    Location
    The Knights Who Say Ni!
    Posts
    27
    I think I qualify as a stat-ho (at least I have been called such).

    That said, I am fine with the plan.

    Insidious does bring up a good point though. Do you see a big difference in overall scoring using the new algorithm? If there is going to be a big difference, I am not opposed to zeroing out the stats and starting over. Phase 1, Phase 2, etc. Afterall, the purpose is to the refine the algorithm.

  10. #10
    Howard,

    I think you may want to have gen 0 worth some points, perhaps just to get new people started.

    I don't know maybe 10 points or something?

    Just a question, how much of a difference are we talking in terms of folding time? From your suggested scoring it seems that gen 2 folds take 4 times as long as gen 1 folds. In which case the gen 0 folds would be relatively instant to the rest of them.

    Or perhaps you want to reward people more for folds of more advanced generations, and the scoring is not representative of the times?

    In any case, I think the ability to work around the last two items is crucial.

    And I would definitely be willing to help out in a beta for the new algorithm. If we're producing better results than I couldn't be happier.
    Team Anandtech DF!

  11. #11
    25/25Mbit is nearly enough :p pointwood's Avatar
    Join Date
    Dec 2001
    Location
    Denmark
    Posts
    831
    Originally posted by Insidious
    I also am wondering how the new scoring system will correlate to the scoring system that is in place at present. Is this new system going to make it impossible to close large gaps in scores created when the clients were fast? (ie: will newcommers have no chance of ever catching up?) That might cause a bit of heartburn for up & commers and/or discourage new participants from joining up.

    -Sid
    Wouldn't that make it unfair to the teams that have a lead? How would you make it? Besides, all you need is to win in the daily stats and eventually, you'll be #1

    The only possible solution I can see, would be to close this "version", announce a winner and start from scratch.
    Pointwood
    Jabber ID: pointwood@jabber.shd.dk
    irc.arstechnica.com, #distributed

  12. #12
    The beta-testing of the screen saver was sort of odd. Needing hard core, early adopters, but testing something they wouldn't use.

    If there is an open beta for the new algorithm, I will do what I can to publicize it on the TSF page and over at Ars to get some technically knowledgeable and thorough testers involved. I am sure the others teams would do so as well at thier respective homes/teams.

    The last two items are the big problems, as I see it. However, that is why you brought this up now - so we can point out problems or circumstances that you have no reason to have experience with or to have thought of.

  13. #13
    25/25Mbit is nearly enough :p pointwood's Avatar
    Join Date
    Dec 2001
    Location
    Denmark
    Posts
    831
    Originally posted by Brian the Fist
    Actually, that is exactly what we intend to do. Although based on the dismal turnout for the screensaver 'beta test', i don't know how well it will work.
    I think that's primarily because most people have no interest in the screensaver. I have very briefly tried it on 2 machines and didn't have any problems. I bet you'll see a very different kind of interest in a beta release of a new console client.
    Anyhow yes, we should be able to get the sneakernet working fine.
    That I think is a very important thing. Having that feature and mking it as easy as possible is critical to many IMHO.
    Also I forgot to mention, stats on Team pages will also list total for current protein and overall so newcomers will still have something to show off (but top 10 will still be overall - I might add a top 10 for current protein as well though, we'll see).
    More stats are always welcome
    The scoring will be as fair as possible and have thought of and will implement ways to discourage people from trying to, for example, repeatedly generate and upload generation 49 structures (which are worth more). You could keep generating generation 0 structures (which are faster to generate that all others) but this is why you'll get less points for them - maybe even none - they are just the 'stepping stone' required to get into generation one where the real work begins. Generation 0 structures are like what you're making right now.

    Actually I like that. How would people feel if you got zero points for gen. 0 and then got scaled points for future generations as mentioned above? It takes only about 1/2hr-2 hrs. to make the first 5000 for most people, and then days to make the rest, so you wouldn't get credit for that couple hours work, but it would be required to reach gen. 1 where you DO start getting credit. I think this is fair and will encourage people to get up high in the generations. Ok, where are you stats-ho's? Does this sound reasonable?
    I have no problem with that. As long as it is equal to everyone I don't see any problems with that.

    Come to think about it - could you make it an option to make that a benchmark? It would be cool if the client had a standard way to make a benchmark. Or maybe if the client had a little benchmark protein (which should, of course, always be the same) and you then made it possible to generate x numbers of structures based on that protein with "foldtrajlite.exe -benchmark" or something?

    That would make it much easier to compare CPU's, OS'es, etc.

    EDIT: Of course, the client shouldn't submit any data to the server when using the benchmark protein
    Pointwood
    Jabber ID: pointwood@jabber.shd.dk
    irc.arstechnica.com, #distributed

  14. #14
    Fixer of Broken Things FoBoT's Avatar
    Join Date
    Dec 2001
    Location
    Holden MO
    Posts
    2,137
    Originally posted by Brian the Fist
    so you wouldn't get credit for that couple hours work, but it would be required to reach gen. 1 where you DO start getting credit. I think this is fair and will encourage people to get up high in the generations. Ok, where are you stats-ho's? Does this sound reasonable?
    well, i am only a moderate stats whore

    but i think that might be a good way to handle it
    Use the right tool for the right job!

  15. #15
    I like that benchmark idea!
    Team Anandtech DF!

  16. #16
    25/25Mbit is nearly enough :p pointwood's Avatar
    Join Date
    Dec 2001
    Location
    Denmark
    Posts
    831
    Originally posted by FoBoT
    well, i am only a moderate stats whore
    I have a very difficult time believing that FoBot
    Pointwood
    Jabber ID: pointwood@jabber.shd.dk
    irc.arstechnica.com, #distributed

  17. #17
    Senior Member
    Join Date
    Jan 2002
    Location
    England, near Europe
    Posts
    211
    If there is going to be a change in the way work is "scored", I believe the only fair thing to do is re-zero all the team stats and start again from phase 2. Announce the winners for phase 1 (top 100 teams, top 1000 users etc) and start afresh. This way, everybody will be starting on a level playing field and new users/ teams will feel that they have a chance to compete.

    I remember when stanford changed the scoring for G@H work units and it upset a lot of members.
    Train hard, fight easy


  18. #18
    Originally posted by Brian the Fist
    Actually, that is exactly what we intend to do. Although based on the dismal turnout for the screensaver 'beta test', i don't know how well it will work.
    As other people said, most of the hardcore people don't run the screensaver so they didn't want to bother trying to install it and test it. I will pre-volunteer for beta testing the new version (especially since I might have to make changes to dfGUI to accomodate it). You can sign me up now.

    What you have mentioned so far including your suggestion for stats sounds fine to me. As long as in your original post the last 2 points making it more difficult to copy to machines and nonetting are scratched everything else looks good. I especially like that you will be posting totals and current protein stats as well.

    Looking forward to this new version.

    Jeff.

  19. #19
    OK m0ti, my attempts to google for it and to look at your team's web site didn't lead me anywhere so I guess I will just ask -

    What is dfQ?

  20. #20
    dismembered Scoofy12's Avatar
    Join Date
    Apr 2002
    Location
    Between keyboard and chair
    Posts
    608
    I don't really think it will be necessary to zero out all the stats. after all, the relative sizes of the proteins have been changing with every update anyway, so each has been effectively worth a different number of points. i think it is fair to assign points to each generation relative to the time it takes to generate, maybe normalized to a recent protein and leave it at that. i think resetting the stats altogether would be much more disruptive than just continuing, especially since there is already the precedent that not all proteins are equal anyway.

  21. #21
    Fixer of Broken Things FoBoT's Avatar
    Join Date
    Dec 2001
    Location
    Holden MO
    Posts
    2,137
    Originally posted by TheOtherPhil
    If there is going to be a change in the way work is "scored", I believe the only fair thing to do is re-zero all the team stats and start again from phase 2. Announce the winners for phase 1 (top 100 teams, top 1000 users etc) and start afresh. This way, everybody will be starting on a level playing field and new users/ teams will feel that they have a chance to compete.

    I remember when stanford changed the scoring for G@H work units and it upset a lot of members.
    if the scoring was frozen and restarted as a new phase, it would make deciding on the new scoring much easier, since the whole angle of trying to keep it on par with the old stuff wouldn't matter

    hmmm, that is something to consider
    Use the right tool for the right job!

  22. #22
    Fixer of Broken Things FoBoT's Avatar
    Join Date
    Dec 2001
    Location
    Holden MO
    Posts
    2,137
    Originally posted by MAD-ness
    OK m0ti, my attempts to google for it and to look at your team's web site didn't lead me anywhere so I guess I will just ask -

    What is dfQ?
    it is still in the pre-alpha stage i believe

    his plan is to write a program similar to SETI Que
    a third party cacheing proxy so that people with many PC's that aren't internet connected (the sneakernet crowd) can have thier non-internet PC's connect to a single point/proxy to send/recieve work. that single dfQ/proxy machine would then be the only machine that would be required to connect (somehow) to the internet/back to DF HQ (Howard) to send results back to the project
    Use the right tool for the right job!

  23. #23
    Yes, FoBoT, thanks for the answer. It will allow a drastic cut-down in sneaker-netting. If you've got a single machine connected to the internet in a lan then you won't have to sneaker-net. Otherwise, you can collect all of the results (in a single file) from a single machine, transfer them to a machine with an internet connection and upload them.

    It's been an off and on project for a while, depending on how utterly insane my life is.

    It's pre-alpha and I haven't worked on it in a while since I'm trying to complete some crunching (about 2 - 3 days worth after some optimizing... and then some more optimizing...) which I then have to analyze for a paper I want to put out (non-DF related).

    In the meantime, there's always dfDetect which is a little Win32 utility I made (by team-mate request's) that does some worthwhile stuff:

    - works for DF installed as a service (or multiple services) or CLI.
    - make sure DF is always running
    - run DF in completely hidden mode (useful for Win9x users)
    - restart DF after X minutes have passed
    - stop DF/keep DF from running while program X is running (great for corporate farmers with CAD machines/gamers).

    In any case, I'll probably be back doing the dfQ thing soon (I hope). February perhaps will see a relase (I was originally hoping for January, but, life got in the way ).
    Team Anandtech DF!

  24. #24
    Note that your dfQ may have to be changed a bit when the new algorithm is released.

    Anyhow, Im going to hold you all to those beta testing promises!

    Now it appears there's 2 camps. Some say reset the stats to zero, some say leave them. Personally, I don't care about stats as you know so I'll do whatever sounds most reasonable.

    To help you decide, a bit more about the 'new' scoring. Each generation should on average take about the same time to generate (except the gen. 0 which is fast and gets zero points anyways). The time will still increase with protein size but not as much necessarily. You will get higher points for later generations (gen. x ponts = sqrt(x) * gen. 1 points) to encourage and reward you for running it long enough to get to higher generations.
    We will set the points for gen.1 to roughly correspond to the number of structures of length 100-150 that you could generate with the present method in the average expected time it will take to finish gen.1 (which should be roughly CPU independent if you read that sentence carefully). This number is yet to be determined and depends on the final generation size we choose but figure around 1000-10000.

    So opinions? Reset stats to zero or keep 'em? I'm especially interested in hearing from people in the top 10 on this, who might be the most pissed off if stats are zeroed...
    Howard Feldman

  25. #25
    Senior Member KWSN_Millennium2001Guy's Avatar
    Join Date
    Mar 2002
    Location
    Worked 2 years in Aliso Viejo, CA
    Posts
    205
    A top ten-ner checking in.

    1. I am okay with resetting the stats and starting the competition from zero on Phase II since it IS a fundamentally new algorithm. If handled well this could be an opportunity to generate an influx of new participants.

    2. I am also okay with leaving the stats as they are now and just adding the new stats to the totals already generated. Since the stats calculations are changing for everybody at the same time it is essentially the same as the fast protein/slow protein differences that we have seen up to this point.

    3. My gut feel is that resetting the stats to zero and calling it DF phase II would be the most fair to all concerned. I am excited that we are trying new algorithms in an attempt to improve the quality of the science.

    .. a quick note to those who are wondering why my account isn't generating the huge numbers that it once did... be assured that my machines are still contributing just as much as they always have to DF, they are just going into different buckets.

    also... lemonsqzz, read your pm.

    Ni!
    Last edited by KWSN_Millennium2001Guy; 01-16-2003 at 10:49 AM.

  26. #26
    Fixer of Broken Things FoBoT's Avatar
    Join Date
    Dec 2001
    Location
    Holden MO
    Posts
    2,137
    Originally posted by Brian the Fist

    So opinions? Reset stats to zero or keep 'em? I'm especially interested in hearing from people in the top 10 on this, who might be the most pissed off if stats are zeroed...
    well, another way to look at this, is not who you will loose , but who you won't gain

    ie

    if you reset the stats, you take the risk of pissing off current crunchers, that is, you may drop in active participants

    if you leave the stats the same, but new people have less chance of catching up, due to differences in the new scoring under the new algorithm, then it is likely that you keep the majority of the old crunchers, but maybe new people look at the numbers and decide to pass on joining DF, because they feel they have no chance to move up

    what i mean is, i think there is more short term down side risk to resetting the stats than to keeping them and adding to them with a new scoring system under the new algorithm/system

    so the "safe" bet is to leave things alone as much as possible, but by doing this, you may be limiting the future potential gains of an influx of new peoples

    however, i am not afraid of a shake up myself, so
    Use the right tool for the right job!

  27. #27
    dismembered Scoofy12's Avatar
    Join Date
    Apr 2002
    Location
    Between keyboard and chair
    Posts
    608
    This may have been mentioned, but it doesn't seem like it would be difficult to keep two sets of stats, a phase 2 set, and an overall set. as to which one would be more "official" or "valid" could be left up to debate

  28. #28
    Member
    Join Date
    Jun 2002
    Location
    Virginia
    Posts
    40
    I don't have a problem with re-setting to zero. It evens the playing field for all teams and team members so there is a big plus for that if you were a latecomer to the program.

    If it makes it easier for your purposes, then by all means simplify it and zero it out.

  29. #29
    Member
    Join Date
    Dec 2001
    Location
    Oslo- Norway
    Posts
    96
    I'll go for two sets of stats, overall and phase 2, IF that's OK with the statsgurus.
    Terje Larsen

  30. #30
    The new algorithim sounds great. I'm all for zeroing out the stats. We have a few inactives with a larger total than me that I could catch a lot quicker!
    TeAm Anandtech Distributed Computing

  31. #31
    Junior Member
    Join Date
    Apr 2002
    Location
    The Knights Who Say Ni!
    Posts
    27
    top ten-er (barely...and probably for not much longer---there is some serious horsepower closing on me quickly!)

    I am fine with any of options: Reset to zero, continue on, or start a new set of stats. It's all good. I just appeciate the openess and dialog.

  32. #32
    Senior Member
    Join Date
    Apr 2002
    Location
    Santa Barbara CA
    Posts
    355
    I think that either zeroing the stats and starting a new phase is fine, or continuing on with current stats. The WUs generated so far have had enough variety in how long they took to make that if the new system is off by some it doesn't matter that much.

    Around one year seems to be a reasonble length for a phase of this project. If you think that there will be more future big changes like this you should probably start a new phase. The next time it seems like it would be harder to zero the stats.

    As I already spend too much time checking the stats I think having total and phase stats would be a little too much.

    Finding beta testers for a new real client should be way easier than a windows screensaver.

  33. #33
    I personally don't have a problem with zeroing out the stats and starting over, but I really do fear what it would do to the project. People who haven't been running the project from the begining would feel that their month or two of work has been a waste of time, and those who are obsessed with stats...well, I don't even want to think about that

    As has already been said, we've changed protiens, tweaked the client to produce over twice as many structures in the same amount of time, and made other changes, and none of them have warranted any changes to the stats. Each protien has varied in the amount of individual production, the new client should be no different, as long as everyone has to do the same thing. Say, no credit unless you finish the entire run.

    Ni!
    Oh, what sad times are these when passing ruffians can say Ni at will to old ladies..

  34. #34
    Ancient Programmer Paratima's Avatar
    Join Date
    Dec 2001
    Location
    West Central Florida
    Posts
    3,296
    So, after a year (!), I FINALLY make it into the top 20 ( ) and you want to flush it all away!!!

    Well, flush 'em. I was considering starting over from zero with a different name, anyway.

    The journey is way more fun than the destination.

  35. #35
    Junior Member
    Join Date
    Aug 2002
    Location
    Palm Harbor, FL -- USA
    Posts
    7
    I'm rather new to the DF project but IMO what should be done is what's best for the project. I have to admit that I'm a stats whore but I understand the goal of the project is more important than the stats.

    Sooo... I'm fine with resetting them to zero, keeping dual stats, or modifying the stats once the new (slower) program is in place. It does sound like the easiest thing would be to reset the stats and start from scratch though.

  36. #36
    Member vsemaska's Avatar
    Join Date
    Mar 2002
    Location
    Connecticut, USA
    Posts
    49
    1) Don't worry about finding beta testers. The dismal response to the ScreenSaver was because most of us are hardcore crunchers and had no interest in it. A new algorithm that may produce better science is a totally different story.

    2) My vote is to zero the stats. When F@H went from V1 to V2 they didn't carry over the stats and there were lots of complaints. Gradually all that died down and it was back to business as usual.

    Only suggestion I'd make is keep the Phase I stats around somewhere in case people want to see them. Maybe www.statsman.org could do something.

    Vic

  37. #37
    I am all for resetting the stats - the people at the top got there by having the best systems and most time commitment - this is something that will not change and they will have no problem staying on the top. Besides, doesn't it get boring being so far ahead of the competition?

    The people at the bottom will be happy to have a chance to catch up, and to get a more accurate idea of how they stack up.

    I would definitely like to see more short-term stats in the future, such as highest ranked per protien, per week, per day, etc. This would let new-comers feel an immediate impact.

    -OgreChow

  38. #38
    Small Time Cruncher
    Join Date
    Oct 2002
    Location
    Away down South in Dixie
    Posts
    28
    I'm not sure I fit in the category of stats-ho or not, but.... After all, I did set up a one-person team because being 198 is better than being 1000+.

    My druthers would be to reset the stats. Most of the folks at the top got there with some mad horsepower, and they'll probably jump out in front again pretty quick. Plus resetting the numbers would get the inactive folks off the list, which might be a positive to some folks (those obsessed with stats).

    It also seems that the new algorithm results might not be suited for a straight 'results returned' number, but more of a points scenario. On the united devices Think projects, the results are tracked by CPU time, results returned, and points based on power, etc. Something similar might work well - consider the sets returned at each generation as a results set, and count them for each generation. Then have a points calculation based on generation, time to complete, CPU speed, kilowatts used, or whatever.

    But hey, as long as I can watch the pretty pictures every once and awhile...

  39. #39
    Senior Member
    Join Date
    Mar 2002
    Location
    MI, U.S.
    Posts
    697
    Yeah, the pretty ASCII pictures are all I want to see, really.

    If you couldn't tell, I don't care one way or the other what you do with resetting. If you reset them, then it might be a better indication of how we're doing with respect to who's running the client at that instant, but that kind of info will go out of date really fast anyway. If you don't reset them, meh, status quo.

    *shrug* Do whatever.

  40. #40
    It seems obvious that those with a lead would be the ones with the greatest motivation to remain the status quo.

    In my opinion, as a member of TSF, if the stats were zeroed it would be real nice to see a page where the final rankings of "Phase 1" were recording. Not only for TSF but for the other teams who have worked so hard to achieve thier rankings and the same goes for individual rankings.

    That said, I believe that having a fresh start (but carrying the current momentum of the project as well as many of the more robust features and stats that have developed over the last year) could be a GREAT boon for the project, especially encouraging those who are intimidated by the incredibly high scores recorded by some teams and individuals.

    If you put the amount of logical thought into this situation as you have most of the other situations I don't foresee a large problem.

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
  •