Actually in your example

1st points =(1000 / 10000) * 10 = 1
2nd points = (1000000 / 10000000000) * 100000 = 10

so no they wouldn't be higher rated.

It's simply a fun way to attempt to grade overall *contribution* across
projects.

Agreed that it can give small projects more points still at least whilst
they have a very small subset of users where that includes 1 or 2 powerhouse
hitters but because of projects like MilkyWay and Aqua giving it ridiculous
amounts of credits thereby basically making the current combined credits
meaningless I think it's a better reflection of contribution right now at
least in a leaderboard sense which is really the aim.

Phil



On Tue, Jul 28, 2009 at 3:20 PM, <[email protected]> wrote:

> I don't understand the basis of the scoring system you are proposing.  This
> still looks like it gives small projects a share of the ranking that is too
> large.
>
> A person that has 1000 points in a project with 10 users and 10000 points
> total would be higher rated than a user that has 1000000 points in a
> project that has 10000000000 points and 100000 users.
>
> If points are equal across projects, this weighting seems absurd.
>
> jm7
>
>
>
>             Phil
>             <freedc....@gmail
>             .com>                                                      To
>             Sent by:                  BOINC Dev Mailing List
>             boinc_dev-bounces         <[email protected]>,
>             @ssl.berkeley.edu         boinc_stats
>                                       <[email protected]>
>                                                                        cc
>             07/28/2009 02:55
>             PM                                                    Subject
>                                        Re: [boinc_dev] credit goals
>
>
>
>
>
>
>
>
>
>
> I've recently been playing about with various simple algorithmns within the
> stats to see what various criteria once summed up show in place of combined
> credits.
>
> In an attempt to offset individual projects credits swallowing up smaller
> projects I've based it (for now) as
>
> points = (user credits / total credits in project) * total users in project
>
> After running this algorithm through my database and then summing for total
> 'points' it gives the top 20 users as follows (rankpoints2)
>
> I include the data for their current total combined points (metric1) and
> another points scoring mechanism  (rankpoints1) purely based upon relative
> rank positions exactly how they do it at the DC-Vault. However I think the
> rankpoint2 scoring works better.
>
> It clearly shows the massive contribution to boinc of the german grid on
> Einstein (13% of the total points there)
>
> mysql> select nick,metric1,rankpoints1,rankpoints2 from
> temp_boinc_combined_user order by rankpoints2 desc limit
> 0,50;
>
> +----------------------------------------------------+------------------+------------------+------------------+
>
> | nick                                               | metric1          |
> rankpoints1      | rankpoints2      |
>
> +----------------------------------------------------+------------------+------------------+------------------+
>
> | AEI eScience group, for the German Grid (D-Grid) a | 1071903064.39851
> |            10000 | 32699.4388545174 |
> | NEZ                                                | 671362889.708666 |
> 80355.726367351 | 26719.5993298215 |
> | ATLAS AEI Hannover                                 | 461412150.282866 |
> 9999.957399495 | 14075.8235478806 |
> | Steffen Grunewald, for Merlin/Morgane              | 419668400.847795 |
> 9999.91479899 | 12802.3901306748 |
> | mac2312                                            | 281136536.678986 |
> 159532.108310697 | 8466.61057314075 |
> | lunarcom                                           |  163324557.86707 |
> 169233.120347095 | 8173.88454355855 |
> | zeitgeistmovie.com                                 | 412503121.525896 |
> 739522.103081642 | 7629.84273592259 |
> | Karl Catling                                       |   10105689.37412
> |            10000 | 6575.17002620478 |
> | Chimichango                                        | 163238495.017144 |
> 59903.27117881 | 5302.35061381748 |
> | Bruce Allen                                        | 170849007.936546 |
> 9999.872198485 |  5211.9140936601 |
> | MikeC                                              |  147704508.38012 |
> 162079.909670802 | 4908.28368902193 |
> | Nebula                                             | 144114223.753479
> |            10000 | 4856.88868657759 |
> | utg                                                | 104967002.416253 |
> 68188.732256869 | 4622.83804779254 |
> | UW-Madison CAE                                     | 147261385.533217 |
> 33822.151402108 | 4606.06700524372 |
> | clemsontiger                                       | 191995966.807699 |
> 37935.420769229 | 4454.18578665772 |
> | Bog                                                | 127516784.375437 |
> 19998.361107242 | 4220.18375769962 |
> | jeidler                                            | 102979934.410189 |
> 348552.997514108 | 4093.34607788475 |
> | GercoG                                             |  42652341.701247 |
> 378454.16821373 | 3800.19507040453 |
> | [B^S] Ralfy                                        |  60357153.021994 |
> 716523.163413209 | 3760.44001449661 |
> | PoorBoy                                            | 193699898.217796 |
> 736468.906205361 |   3532.549369718 |
>
> I'd welcome any discussion on this, not sure if it would be feasible to
> work
> into the BOINC framework itself..
>
> Phil
>
>
> On Fri, Jul 24, 2009 at 11:30 AM, <[email protected]> wrote:
>
> > The project should only include computers that actually report the
> average
> > in calculating the correction.
> >
> > jm7
> >
> >
> >
> >             "Lynn W. Taylor"
> >             <[email protected]>
> >              Sent by:
> To
> >              boinc_dev-bounces         [email protected]
> >             @ssl.berkeley.edu
>  cc
> >                                       BOINC dev
> >                                        <[email protected]>
> >              07/24/2009 11:04
> Subject
> >              AM                        Re: [boinc_dev] credit goals
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > ... but only if the client participates in more than one project???
> >
> > Otherwise, you have a whole bunch of clients reporting the average for
> > the project that needs to adjust.
> >
> > [email protected] wrote:
> >
> > > One possibility is that every client tracks credits granted per hour of
> > CPU
> > > time (and separately GPU time) and sends the average of that to all of
> > the
> > > projects to which it is attached.  The project takes the average of
> > those,
> > > and the average credit grant it makes to determine a correction to be
> > > applied to the next X time worth of credit grants.
> > _______________________________________________
> > boinc_dev mailing list
> > [email protected]
> > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> > To unsubscribe, visit the above URL and
> > (near bottom of page) enter your email address.
> >
> >
> >
> > _______________________________________________
> > boinc_dev mailing list
> > [email protected]
> > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> > To unsubscribe, visit the above URL and
> > (near bottom of page) enter your email address.
> >
>
>
>
> --
> Distributed Computing stats
> http://stats.free-dc.org
> _______________________________________________
> boinc_dev mailing list
> [email protected]
> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
>
>
>
>


-- 
Distributed Computing stats
http://stats.free-dc.org
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to