Overclock.net › Forums › Overclockers Care › Overclock.net Folding@Home Team › Chimp Challenge 2012??
New Posts  All Forums:Forum Nav:

Chimp Challenge 2012?? - Page 5

post #41 of 314
Its called the finer detail man. What, do you think this stuff just happens?

This is the detail that is usually hammered out in some back room between the team capts. Im liking the idea of seeing what goes into the final comp.

Yeah, its for fun and its for research but this needs to happen and Im glad its happening where everyone can see it.

If your not interested, fine but dont go throwning hand grenades into this post because the detail is too much for you.
Edited by Sethy666 - 4/12/12 at 11:38pm
post #42 of 314
if we got rid of using Chimp names then like I said EVGA might take some Hit for a fixed Handicap
look at the avg gap 4M ppd isn't nothing - we just might handicap by some margin that would make it close between us

however as Adak said there is nothing barring your team from getting another team to join you and passing us quite easy if you had an early lead time and invited a team to join you
then padding or handicap would not be necessary - in fact EVGA would be stuck having to do the same thing just to overcome the merger (or really have on spectacular membership drive)


you are the smaller team you are suppose to overcome adapt and get new folders to beat us.... that's the challenge in the PPD
that's what the race is for to grow the teams and PPD in a way for you to win?

4M PPD - 1 user on HPCS did 2M PPD - actually 2 users did that at EVGA (similar system)
10 users at 100k =1M that's 40 users - do-able - you will gets some 4p or 2p and smaller system - but it will even out - it might take 60 - even 80

if it was 2M that's 20-40 more new folders...adapt overcome - you could out pace us at new users and beat us with a gap that small

it isn't to make a formula that take away any benefit of being the top folder and the target to beat and turn them into a player for last like last year

we should have some advantage - and the challenge should be to push others to beat us after all
but we shouldn't be smoking you out of the water

evga looked at #1 and #2 ideas before adak these are normal handicaps -
however if teams did join and had massive over production a timed race would leave little time for a team to adapt and overcome the deficit

and with #2 this is what EVGA considers with OCN for fair race - the others would be out of it for the top spot

however I wouldn't mind a fixed Point for each of them with a close gap to make them also work for it
like 10M PPD which would put you guys pretty close to OCN numbers (but not beat us or them - because after all you can join teams and blow us out of the water)
if you want variance you can add/more to equal the other teams out a bit more too

like put OCN 1 M off our PPD and everyone else 1 M off of ocn (with some close variance)..if you want it tighter it can be discussed

EVGA will probably be not joining up with teams so anything like this could kill us or OCN in the PPD. where smaller teams would again benefit - hence they are to be set back a bit - and not even with us
as long as they are relatively small teams joining together it be a race - like 8M team +10M handicap then you add a 4M -6M team it be close

if the gain is too great then the handicap fails again ..because big teams won't merge teams or invite teams to join them normally

but this be for the PPD race only


in handicap - both EVGA and OCN are at a disadvantage with just percent growth - teams joining up or new users would be the biggest issue there that's one reason why EVGA went so low in the last race the other was conversion (this was an added penalty)

and handicap like that hurts more - and then a bad handicap formula delivered to the team before the race - kills any chance of getting the team out
it's a double whammy...if the team perceives it as bad,and it's analyzed to really be bad (we would run numbers just like any other team)

then we wouldn't get the team out... after all the handicap reduces ability to overcome and only favors the weakest to easily overcome any threat

at least by removing the CC names this become less of an issue - and the people that wouldn't have joined are now included regardless if they wanted out (to opt out they have to completely stop folding)
because most would have folded anyways...but not for the CC


so like you said the only way is to run both - one for each type of team


you could also analyze the team data after to see if we could come up with an overall between the two and figure a balance point where a certain % of PPD weight and % of growth to come up with something



but we would have to see how the teams fair under both with full out races in each - to see how they perform -
if you apply it early and the percentages are wrong - you may have teams question what it would do and may fold less (because they know it's not fair again)
we need a good baseline of each by themselves - then apply a formula on top after we look at the data

like 45% weight on PPD and 45 on PPD growth and 10% on another factor like membership growth
(you would have to work this out after you had unbiased data)

otherwise we are chasing a snake for a few years as it wiggles each way- back & forth (one side wins then the other..back&forth - till we get it right)
post #43 of 314
I think the added burden of making/reminding people to change names is not what the competition is about and is just another layer for confusion or mistakes etc. The competition is about team folding not reminding people to be sure to put their the new cover sheets on their TPS reports. So I'm for just the teams being counted as the teams, instead of usernames. We can still call ourselves OCNChimpin' though.
Edited by juano - 4/13/12 at 12:36am
Main Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7 2700k ASUS P8P67 WS Revolution EVGA 980 Ti SC+ Samsung 4x4GB DDR3 1866MHz 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 1TB Samsung Spinpoint F4 2TB Samsung BD Combo Noctua NH-D14 
OSMonitorPowerCase
Windows 10 64 bit Asus PG279Q Kingwin Lazer Platinum 1000W Silverstone Raven RV03 
  hide details  
Reply
Main Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7 2700k ASUS P8P67 WS Revolution EVGA 980 Ti SC+ Samsung 4x4GB DDR3 1866MHz 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 1TB Samsung Spinpoint F4 2TB Samsung BD Combo Noctua NH-D14 
OSMonitorPowerCase
Windows 10 64 bit Asus PG279Q Kingwin Lazer Platinum 1000W Silverstone Raven RV03 
  hide details  
Reply
post #44 of 314
Quote:
Originally Posted by Xavier Zepherious View Post

if we got rid of using Chimp names then like I said EVGA might take some Hit for a fixed Handicap
look at the avg gap 4M ppd isn't nothing - we just might handicap by some margin that would make it close between us
however as Adak said there is nothing barring your team from getting another team to join you and passing us quite easy if you had an early lead time and invited a team to join you
then padding or handicap would not be necessary - in fact EVGA would be stuck having to do the same thing just to overcome the merger (or really have on spectacular membership drive)
you are the smaller team you are suppose to overcome adapt and get new folders to beat us.... that's the challenge in the PPD
that's what the race is for to grow the teams and PPD in a way for you to win?
4M PPD - 1 user on HPCS did 2M PPD - actually 2 users did that at EVGA (similar system)
10 users at 100k =1M that's 40 users - do-able - you will gets some 4p or 2p and smaller system - but it will even out - it might take 60 - even 80
if it was 2M that's 20-40 more new folders...adapt overcome - you could out pace us at new users and beat us with a gap that small
it isn't to make a formula that take away any benefit of being the top folder and the target to beat and turn them into a player for last like last year
we should have some advantage - and the challenge should be to push others to beat us after all
but we shouldn't be smoking you out of the water
evga looked at #1 and #2 ideas before adak these are normal handicaps -
however if teams did join and had massive over production a timed race would leave little time for a team to adapt and overcome the deficit
and with #2 this is what EVGA considers with OCN for fair race - the others would be out of it for the top spot

however I wouldn't mind a fixed Point for each of them with a close gap to make them also work for it
like 10M PPD which would put you guys pretty close to OCN numbers (but not beat us or them - because after all you can join teams and blow us out of the water)
if you want variance you can add/more to equal the other teams out a bit more too

like put OCN 1 M off our PPD and everyone else 1 M off of ocn (with some close variance)..if you want it tighter it can be discussed
EVGA will probably be not joining up with teams so anything like this could kill us or OCN in the PPD. where smaller teams would again benefit - hence they are to be set back a bit - and not even with us
as long as they are relatively small teams joining together it be a race - like 8M team +10M handicap then you add a 4M -6M team it be close
if the gain is too great then the handicap fails again ..because big teams won't merge teams or invite teams to join them normally
but this be for the PPD race only
in handicap - both EVGA and OCN are at a disadvantage with just percent growth - teams joining up or new users would be the biggest issue there that's one reason why EVGA went so low in the last race the other was conversion (this was an added penalty)
and handicap like that hurts more - and then a bad handicap formula delivered to the team before the race - kills any chance of getting the team out
it's a double whammy...if the team perceives it as bad,and it's analyzed to really be bad (we would run numbers just like any other team)
then we wouldn't get the team out... after all the handicap reduces ability to overcome and only favors the weakest to easily overcome any threat
at least by removing the CC names this become less of an issue - and the people that wouldn't have joined are now included regardless if they wanted out (to opt out they have to completely stop folding)
because most would have folded anyways...but not for the CC
so like you said the only way is to run both - one for each type of team
you could also analyze the team data after to see if we could come up with an overall between the two and figure a balance point where a certain % of PPD weight and % of growth to come up with something
but we would have to see how the teams fair under both with full out races in each - to see how they perform -
if you apply it early and the percentages are wrong - you may have teams question what it would do and may fold less (because they know it's not fair again)
we need a good baseline of each by themselves - then apply a formula on top after we look at the data
like 45% weight on PPD and 45 on PPD growth and 10% on another factor like membership growth
(you would have to work this out after you had unbiased data)
otherwise we are chasing a snake for a few years as it wiggles each way- back & forth (one side wins then the other..back&forth - till we get it right)

I have no problems with you. I also have no say in this. Just gathering info. My folding name is hertz9753 what is yours?
    
CPUMotherboardGraphicsGraphics
Intel i7 2600k ASUS P8Z77WS EVGA GTX 980 SC ACX 2.0 EVGA GTX 980 ACX 2.0 
GraphicsRAMPower
EVGA GTX 970 SC ACX 2.0 Mushkin Enhanced Blackline 8GB (2 x 4GB) 240-Pi... FirePower ZX Series 1250W 80Plus Gold 
  hide details  
Reply
    
CPUMotherboardGraphicsGraphics
Intel i7 2600k ASUS P8Z77WS EVGA GTX 980 SC ACX 2.0 EVGA GTX 980 ACX 2.0 
GraphicsRAMPower
EVGA GTX 970 SC ACX 2.0 Mushkin Enhanced Blackline 8GB (2 x 4GB) 240-Pi... FirePower ZX Series 1250W 80Plus Gold 
  hide details  
Reply
post #45 of 314
Quote:
Originally Posted by hertz9753 View Post

Edited
For this example, I just used the highest point value from 2011 CC * 2.

Yes!

But to handicap it right, you'd use actual month folding figures, because lots of folders from teams, did not sign up for the CC last year. If we used the CC data for this, it would throw the race off hugely into EVGA's lap, because their participation level was so low last year.

It's the overall point and race format that I wanted to introduce, and see what you thought of it.

IMO, this format is exciting, and balanced enough (fair) that it could eliminate the need for two brackets. At first, the smaller teams have this HUGE Boosted lead, but then in the last few days, the larger teams come ROARING back, down the back stretch! biggrin.gif

It WILL be difficult to properly set the Boost (handicap the race), because some teams have earned so many points with HPCS, (which is diabolically ending during the race), and others didn't use HPCS much at all, also, some teams had high participation in the CC last year, and others had very low participation.

So handicapping the race right, is a bit of an art. But this is a format that readily can be tweaked - once you use it, you'll understand it fully, and won't need college trig or better, to do it.
Edited by Adak - 4/13/12 at 12:41am
post #46 of 314
I did like that one. smile.gif Let us know on the on the pick. smile.gif
    
CPUMotherboardGraphicsGraphics
Intel i7 2600k ASUS P8Z77WS EVGA GTX 980 SC ACX 2.0 EVGA GTX 980 ACX 2.0 
GraphicsRAMPower
EVGA GTX 970 SC ACX 2.0 Mushkin Enhanced Blackline 8GB (2 x 4GB) 240-Pi... FirePower ZX Series 1250W 80Plus Gold 
  hide details  
Reply
    
CPUMotherboardGraphicsGraphics
Intel i7 2600k ASUS P8Z77WS EVGA GTX 980 SC ACX 2.0 EVGA GTX 980 ACX 2.0 
GraphicsRAMPower
EVGA GTX 970 SC ACX 2.0 Mushkin Enhanced Blackline 8GB (2 x 4GB) 240-Pi... FirePower ZX Series 1250W 80Plus Gold 
  hide details  
Reply
post #47 of 314
Quote:
Originally Posted by Bodycount View Post

Like i said PAIN IN THE ASS! what?, are you two writing a book with your posts? learn to sum things up.
You two make it a pain in the ass in regards to to the point system it's ALL for fun ! and research rolleyes.gif
You guys are starting to miss the point rolleyes.gif

Yes, that's heard a great deal before a race. But things change immediately once the race starts. Then, everyone wants to know "why didn't this detail get handled?", and "why didn't you think about this other thing, before?".

Your POV will change the instant that the race starts.

We need to write a book, because we can't talk to each other on the phone, or in person, and time is short. There is a general feeling of dislike for the way the CC was run in the last year, and something better and different, is wanted.

That kind of change doesn't just happen without communication, and lots of it. Either you work for a better race, or you simply won't have the good race that you could have had.

I agree, it's not for everyone, but we're not everyone - we're organizers for a race! thumb.gif

@hertz
Thanks for looking it over, and I will.
Edited by Adak - 4/13/12 at 12:51am
post #48 of 314
Quote:
Originally Posted by hertz9753 View Post

I have no problems with you. I also have no say in this. Just gathering info. My folding name is hertz9753 what is yours?

my folding name is Xavier_Zepherious EVGA CC Capt
http://folding.extremeoverclocking.com/user_summary.php?s=&u=541641

my thx go to all the team members here that have to put up with this to help save or build a better CC
post #49 of 314
Quote:
Originally Posted by Adak View Post

1) Each team WILL ABSOLUTELY have a team name. Don't know where you got that idea from, but it's completely off the mark.

 

Again, this is what you said earlier:

 

Quote:
Originally Posted by Adak View Post

I've made some changes to the recommendations mentioned above:

1) The race would be a team race - no individuals need to change their folding name or passkey. You would race as a team.

 

If an individual doesn't change their Folding name, how do they Fold under the CC name for their team?

 

Quote:
Originally Posted by Adak View Post

2) I did not see your full proposal, or Xaviers. Just a sort description. If you want to discuss it, then let's discuss it. I would need to see some numbers like I posted, using your formula, before I could discuss it in depth.
 

 

That was more directed at Xavier; it is constantly stated that the conversion rate was unfair, but I never said anything was finalised.

 

Quote:
Originally Posted by Adak View Post

3) If you make the basis based on folding "during the year", then you destroy the ability to handle teams that have numerous temporary folders, join into the race. That is unacceptable, since it lead to the problem of the CC last year.

 

Three issues with this:

 

1) You said that a conversion rate would be unfair to use, however, by taking the basis from the previous CC, you're already including conversion. That would presumably remove your objection to the use of a conversion rate?

 

2) I would like to see some explanation as to how this one factor led to the problem we had last year.

 

3) If you are against using numbers from the remainder of the year, how do you account for the changes in production per team each year? As I've said, OCN increased by 100%, while CPC are at the same level. Why should such varying performance over the course of the year make such a difference in a 10 day event?

 

Quote:
Originally Posted by Adak View Post

4) If team A races with team B, and team B increases it's folding by 10%, then team A needs to increase their folding by 10% to keep up, in a handicap race.
Does that not seem fair?

 

Adak, you're a veteran Folder, and have been around the CC long enough that I'm not gonna talk as if you don't know anything. However, in this situation, you are wrong.

 

This was the exact comment the captains were making last year. I know... I was one of them. I fully believed - as you do now - that the system was fair and would work out well. It didn't. I take my share of the blame from last year's mess;  things didn't work out well, and that was my as much my fault as anyone else's. I can't sit here and accept a system that let's the same mistake be made again. The fact is, for a smaller team, you could bring in one mega-Folder, with 500k+ PPD from another team, and you're automatically the front-runner. If that was the only way to do things, instead of CC teams trying to grow their teams from within, you'd have members going to non-CC teams, trying to entice big Folders away for the 10 days.

 

That is what will happen; for most teams, you're not gonna have one member suddenly increase their PPD by 500k... and most teams are going to struggle to get 10 Folders with 50k PPD for the event too. It's too easy for a small team to make a massive difference. And it's no good for the CC, or the teams themselves, to have to have members scouring other teams for their big Folders, just to be competitive.

 

Quote:
Originally Posted by Adak View Post

5) The race last year had a mistake relating to the HWC. This fixes that problem, easily. But just for a data item, HWC has, WITHOUT NCIX HELP, one of the highest PPD average per folder, of any team in Folding @ Home, (minus the odd supercomputer teams like awachs). So in any handicap race, it would be very hard to keep them out of one of the top finishers, don't you think?

 

I'm not denying that. However, HWC also only have the 20th highest production of all teams in Folding@Home. The two are not separate criteria, and shouldn't be made into discrete categories like you're suggesting - that leaves us with two categories with a team in each almost certain to win, unless they make mistakes. If Evga and HWC compete to their potential, there is very little change for anyone else to win, regardless of how hard they work. I can't speak for other teams, but on OCN, loads of us worked to get everyone to switch over to the CC name... to set up new Folders... to increase our own indivdual PPDs. And the end result? We were still 40% beind the winning team.

 

The same will happen again this year - the bigger teams will try hard to increase their production, but in the end, smaller teams have to do much less to win.

 

Quote:
Originally Posted by Adak View Post

That's why I thought having both a Handicap race bracket AND an Unlimited race bracket, would bring a better balance into the CC. Because handicap races will ALWAYS favor either the little teams (because they're based on percentages), or the biggest teams (because they're based on points alone, even though it's a handicap).

You need both brackets to balance it out.

 

Why do we need two brackets? Why can't we make one system, which factors in both? Handicap-based performance, and absolute performance, and have a fair winner, in all regards? Rather than a big team running away with the first category (since I think we all know only Evga or OCN can win that), and a small team winning the second category (where realistically, one of the smaller teams will outperform the bigger ones), have one category. One which takes into account how well you performance based on you team size, but also how much you produced -  which is actually what is important to Stanford.

Megadoomer
(14 items)
 
Family Computer
(13 items)
 
 
CPUMotherboardGraphicsRAM
Phenom II X6 1090T @ 4.0Ghz ASUS M4A89GTD PRO Sparkle GTS 450 2x4GB G-Skill Sniper 
Hard DriveCoolingOSMonitor
Samsung F1 1TB CM Hyper 212+ Windows 7 Professional x64 Samsung T220 
KeyboardPowerCaseMouse
Logitech MX3000 Laser CM 1000M HAF 922 Logitech VX Revolution 
CPUMotherboardGraphicsRAM
Q6600 Asus PN5-D 750i Evga GTS 250 2x2GB Crucial Ballistix 
Hard DriveOptical DriveOSMonitor
750GB Hitachi Samsung Super Writemaster Windows 7 Professional x64 19" Dell 
PowerCase
Corsair 450VX Antec 900 
  hide details  
Reply
Megadoomer
(14 items)
 
Family Computer
(13 items)
 
 
CPUMotherboardGraphicsRAM
Phenom II X6 1090T @ 4.0Ghz ASUS M4A89GTD PRO Sparkle GTS 450 2x4GB G-Skill Sniper 
Hard DriveCoolingOSMonitor
Samsung F1 1TB CM Hyper 212+ Windows 7 Professional x64 Samsung T220 
KeyboardPowerCaseMouse
Logitech MX3000 Laser CM 1000M HAF 922 Logitech VX Revolution 
CPUMotherboardGraphicsRAM
Q6600 Asus PN5-D 750i Evga GTS 250 2x2GB Crucial Ballistix 
Hard DriveOptical DriveOSMonitor
750GB Hitachi Samsung Super Writemaster Windows 7 Professional x64 19" Dell 
PowerCase
Corsair 450VX Antec 900 
  hide details  
Reply
post #50 of 314
During the CC, OCN points will be given to the team name for the OCN team, in the CC, by the CC race organizers. No one needs to change their folding name, because the CC changes it for them, for the purpose of the race.

Check out the latest format, called "Boost". I believe you will like it. There is no "conversion" factor, like in the last CC. The team's handicap is calculated by the race handicapper, and given to each team as points, at the instant the race starts.

Then the fastest team has to catch the other teams, if it can. It's an exciting format, because it seems like the smaller teams that are folding slower, would have an insurmountable lead. But the faster teams come roaring back in the last few days!

And only one bracket is needed, since it relies on points, and not percentages, already.

Last year, we used percentages. Small teams have an inherent advantage when you use percentages, since they can increase the percentage of their team, much easier than larger teams. We all agree on that. Still, it can be a good race format if it's handicapped right. The problem was that the BASIS for the teams, did not include numbers taken into account when small teams could add several temporary members as teammates - like NCIX members folding with HWC. That's what "blew" up the CC last year.

The change in production during the year is something that the race handicapper takes into account. That will probably be me, and it's a bit of an art. You bring in all the factors, and make estimates based on experience, for the factors that can't be known in advance.

I asked the question, "Does that not seem fair?". I was asking a question. In a good race, "fair" is what everyone feels, is fair - although nothing is absolutely FAIR. That's why I'm asking the question. I want you to have a race format, that you feel is fair, and your team feels is fair, also. That goes for every team and folder.

We don't need two brackets. The Boost format makes one bracket all we need.

We want the teams to bring in new members, for sure. Bringing in folders from other teams is not that helpful to FAH clearly, since they were already folding. I'm not sure how we can stop recruiting from other folding teams, into racing CC teams, without also stopping teams from bringing in new members, as well. I suspect that this is something we will have to accept in the CC, regardless of the rest of the format, if you want new members to be allowed into the CC.

One good thing about the Boost format, is that it's made for handicapping. So it can be tweaked (handicapped), extensively, if it's not just right, from year to year. I'll have to show one or more others how that can be done. All you need is a spreadsheet and a bit of experience.

To appreciate the Boost format, you first need to forget the old format - Boost is very different. Percentages mean nothing in this format - it's all about points, and it's all about catching and passing the teams that have been given a greater boost - without being caught yourself!

I believe you'll like this format, Zodac. smile.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Overclock.net Folding@Home Team
Overclock.net › Forums › Overclockers Care › Overclock.net Folding@Home Team › Chimp Challenge 2012??