AMD Users Competition 2007 - FINISHED

[PoorBoy]

... I didn't even know I was in a race, I'll be glad though when I get a Pharm so I can move up in the Ranks ...

[gamer007]

Gah. Lost top 10 in the last day.

[Steve Lux]

... I didn't even know I was in a race, I'll be glad though when I get a Pharm so I can move up in the Ranks ...

Yeah, perhaps if you're really good and save up all your money you could buy an extra computer in a year or two.

[AMDave]

overall Comp Stats for Race 4 have been updated
(See first post in thread)

Thanks to our Comp-master, NeoGen

Proteins@Home and Folding at Home, here we come.

[NeoGen]

Yea, I updated but forgot to mention.

What I'm really puzzled about is at how much to set the pharmers cut-off line on F@H... there's multi-cores, gpu's and PS3's all mixed on this project. And I don't know the daily average points of those devices.

[moderna]

This is a sign of the times. It is becoming a puzzle.

Tanpaku seemed to generate higher WUs than the other races. I did not do an exhaustive look and I cannot see the computers of some of the pharmers but Skuzz, Anzi and I are single computer pharmers. Skuzz averaged just over 1100! There do not appear to many multi-computer racers in the top non-pharmer group but there are a few.

Do you think there will be many racers running the alternative platforms? If so, the line should probably be high. For Tanpaku it would have been over 1200 if Scuzz could have run an alternative platform. You would have had only 7 pharmers and would have pushed all of the regulars down 7 places with some pharmers moving down the list.

700 would be consistent and we single machine racers can decide if we want to game the number or run full out anyway.

I will continue to play with a handicap idea for the next series.

[Jerod Vandehey]

I have only a single x2-3600 on Tanpaku and I was 7th place overall in the pharmers. It is overclocked 27.5%. Not sure if that helps or not. My RAC is 840. It doesn't bother me to be pharmer or regular. I usually just run the same projects whether or not we are racing.

[moderna]

I have been thinking it would be nice to give more people a chance to win in the races. Races can bring out the best in people as long as they have some base for comparing their performance.

The current race is pretty much geared to total contribution to the team and its position in the project. Throw everything you have against the project to win and you win if you have more or better machines.

Adding some additional categories might be a way to do this without trying to develop a handicap. Handicaps appear to be hard without a great deal of work and more statistics than I can find. Bowling teams use all kinds of categories to equalize things so everyone can compete. Most improved bowler, games over 150 or 200 or something, high series, etc. and of course they have a handicap based on a percentage of the difference of past performance and some target.

I am always unrealistic on this sort of thing but I thought I would give it a shot. The administration is the biggest issue and anything I have come up with makes for more work. I was just thinking it might be fun to open things up for comment. I do not know all of the sources for data so maybe there is a way for something more.

Some ideas that come to mind.

Allow people to enter certain race categories.

This might help to solve the problem of finding a cutoff value for pharmers.

Most improved over last race series.

By total, or by percentage.

Most improved contribution by a single machine.

A racer could dedicate more time to crunching on a machine, or tweek the performance of their machines over the past(?) to give them an edge. That would be like tuning for winning. This would require a baseline for each machine entered and a percentage improvement calculated.

Most improved contribution by a pharm.

Same as above but for multiple machines.

Most tweeked machine.

This would involve comparing actual against a baseline for a particular processor.

Most unexpected contribution.

This would be based on something like a predicted score based on past performance, RAC, WUs in last month, etc. If you recruit a bunch of machines for the race or some other improvement, you win.

Biggest, meanest single computer

Match races, ladders, etc.

One on one challenges. Handicaps could be used.

OK, I'll stop now.

[NeoGen]

I think they're probably all doable ideas, but it needs some study.
One thing I always keep in mind is that the contests are for everyone and not just for us here on the forums, so I can't arrange special conditions that need member interaction/response.

Another thing is that its very complicated to follow individual machines on a contest. On non-boinc projects is next to impossible, there are no records of machines, just teams and members on nearly all projects. On boinc projects it can be done, but many people prefer to keep their machines hidden, which means I can't see everyone's machines.

Apart from those two things, I just try to make it equally fair for everyone... the only resources I have are people's names and points, so I have to make do with it.

But please, do give me some suggestions. I'm always looking for ideas. This year's contest for example was an evolution over last year's. I made the races slightly longer and added new non-boinc races to the mix, because many people still don't fancy boinc, and I felt it wasn't right to leave them out.

[PoorBoy]

There's no real Honest way to run a Contest, with all the Optimized Clients Floating around like v5.50 and others and just simply modifying the .xml File theres to much Fudging going on ... IMO

Also with the Hoarding before the Contest even begins throws another corkscrew into the results, so all it boils down to a lot of times who can Fudge the most and Hoard the Most ...

That's why I have stayed out of the Contests, you just happened to run the Contest at Spinhenge when I had a small Resource Share going there so I got caught up in the Results.

PS: I will say this though, you do a real great job of running the Contest & presenting the Results NeoGen, it was interesting to watch anyway ...