Just an idea

[strates]

I have been considering the problem of ranking all DC teams on a world basis. DC-Vault has done an adequate job using a scoring system based strictly on project rank. But this has it's issues which can be discussed elsewhere. This site, has done an outstanding job of collating stats from many many dc projects, and the DCR section is working nicely (even if I haven't quite figured out exactly how the DCR is calculated).

But, for the last couple of days I have been considering the possibilty of developing a world team ranking system based on the percentage of work a team contributes to the overall project. I would discuss the advantages a feel a ranking system such as this would have, but I would prefer to hear others considerations about the possibility before tainting the water so to speak.

I think the idea is fairly self explanatory, but just in case it is unclear, consider "seventeen or bust" and rosetta@home as an example.

In SoB, Overall work completed (by members both affiliated and non-affeliated with a team) is 144.65 P. Of which 9.06 P was completed by team FreeDC. This is equal to 6.26% for SoB.

In Rosetta, total work is 827,182,917. FreeDC has completed 53,170,475. This is equal to 6.42% for Rosetta.

To get the final ranking value for the team, simply add up the percentages and divide by either total number of projects or just the number of projects the team is participating in, (Both methods should be considered.)

to finish the example....

6.26% + 6.42% = 12.68

12.68 / 2 = 6.34. <--- This number would be FreeDC's world score.

Please consider this proposal thoughtfully, and discuss your feelings on it. I look forward to your comments.

Thanks,
Strates

[PY 222]

I say if you build it, people will flock to see it.

Do you have a page up for this yet? Are you going to track all teams across all projects in the DC world? This is a tall order IMHO.

Or you could do a top 100 teams across the DC World.

[D2DM]

Here a few % on a long time running big project = A few % on a short time running small project.

[Bok]

This is actually fairly similar to my calculation for the Teams DCR except that I only use the combined score of all the teams within the DCR within a project (though this is most of them anyway..)

So for each project I use

Score = ((team score) / (total score of all teams in DCR in project)) * number of participating teams in project.

I could easily produce a version using your scoring method (at least for the projects where I can track all teams that is...).

It would just mean using the total score irrespective of whether the team is in the DCR (which I hold in a table) and then multiplying by percentage rather than number of teams..

Bok

[PCZ]

I personally dont think you can award points like that.

There are projects where you need thousands of computers to do just a few percent of the total.
Other little projects can be totally owned by a small pharmer.

You have to award points based on total computing power used.
The DCR as it stands is useless.

[Bok]

I see.... what is your suggestion then ?

[D2DM]

Suggestion:
Use per project (wu's, cands, blocks, tests, etc.) some sort of measurement like <cpu type> years.

Then
Project A totals 500 <cpu type> years Team one has 10%, team two has 5%
Project B totals 100 <cpu type> years Team one has 5%, team two has 10%
...

So
team one has 0.10x500 + 0.05x100 = 55
team two has 0.05x500 + 0.10x100 = 35
...

strates: your idea would have the two teams equal (my previous post)

[strates]

Bok,

Thanks for clarifying how the DCR is calculated, that actually seems to be a pretty solid method. I don't see any reason including all teams would be an improvement over what you are already doing. I am just wondering if I missed an explanation of this DCR calculation when I was searching through the forums, or is there another place you've stated the calculation method.

D2DM,
I think the problem is that no ranking method would be perfect, but using a percentage based system, does give a valuable snapshot of how much work each team is doing. So what I am saying is that, yes there are other methods, but they have there problems as well. So short of every DC project out there standardizing a workunit size, (much like Boinc has done), there is not going to be a perfect ranking system for all projects.

Thanks,
strates