Punchy

Edit: I forgot to mention that if you ignore the bogus stamina category, then what Gryphon posted in 743 is basically hysterics and incorrect.  Run the numbers for yourself and see the truth.

 
You have no idea what you're talking about. I could try and explain it to you but you won't get it. Either because you're not prepared to accept that you're wrong, or because your math won't suffice. So, I'll really stop here. I'm extremely disappointed, and I'm done with you, and I'm done with this charade. 
 
You think you're out, I'll show you how to be "out". 
 

theGryphon

 
The only outstanding issues I see are:

• How the points are distributed:  Do you want to give teams with an unfair advantage in total production a better standing in the challenge?
  • As I previously stated, we should come up with a system that minimized the different aspects of the diverse teams participating rather than one that accentuates it.

 

 

Such a system exists guys, again it is the one in #671. You may dislike it cosmetically, or offer changes to make the range from 50 to 100 (which is very easy to do), but perfectly balances the categories. 

 
 
Nope.  You're missing the point.  That still favors teams that do significantly better in any particular category. 

well, I'll just wait for an announcement on starting time & client setup, if any.

Punchy, sorry to hear about your dog.

I been reading this every day.WOW whats the big difference who wins or loses.Most of us are going to help EVGA out regardless .
  Its time to **** or get of the pot . I and most that will be folding could care less how points are done as long as it looks ok. STOP nit picking lets just do CC or not do CC.
  At this rate looks like evga crunching is going to be getting my folding comps to.I would hate to just be one more buck folder.Remember PG is really messing up there WU,s.
  Now I may be only a small folder but last I looked most are small folders. A team as large as evga's and in 4 months I am close to being in the top 1000.I am wounder why COULD it be all the bickering over little stuff really can not put it on buck folders . Hell we are on page 26 and no end in sight. So i will say this once more **** or get off the pot
  This is the way I look at it some of us do not have jobs but keep on folding >then there are ones like me 24 hrs a week at min. wage that find a way to fold 24/7 or close to it.Then there are the ones that got my dream computer cost them less to fold then me but make 10x the points.Well WHAT ones you think are bickering sure as hell are not us small Fri.
  So in the end are we doing CC or not. I really would like to know running gpu,s cost to much I got a p67 MB just waiting for ram and a CPU thats like 3 weeks wages for me but what the hell power and interwebz can wait for there cash.
  Just my 2 cents worth on how this is going   .As i hear a t storm coming this way.Hope my power stays on
EDIT That should of been 26 pages and 4 months of nothing

Here's a spreadsheet comparing the 4 most recent formulas I can remember.  I highlighted first-third places picked by each formula for visual convenience.  I can add other formulas if you post them or point to a post where they are documented.

The vote in the captains' forum seems to be in favour of keeping CC names (by a margin of 1 team), so with that in mind, we could keep conversion and not have stamina this year.
 
As for the proposed systems... what is "growth" based off? Points are using last year's numbers, so are you comparing 2011 to current PPD?

Kk, fair enough.
 
Geomean would have been my preferred choice, but the gap from 1st to 2nd is too great... either of the range options seems to be a fair contest (I think the direct points version is too close).

I'm going to write this down, for whatever it's worth, because I'm seeing that Punchy is dragging the system in the wrong direction. 
 
Punchy, you keep bringing up geometric mean, giving Spec.org as an example. What you don't understand is that Spec is applying the geometric mean to combine scores that are very similar to each other. For example, normalized integer-point scores are combined in one score (e.g. CINT2006), normalized floating-point scores are combined in another score (e.g. CFP2006). Integer point scores are measuring the same performance measure, integer performance, only using different algorithms. Therefore, once normalized, the variability of each of those algorithms will be very very close to each other. That is why when you combine the ratios using geometric mean, the affect of one algorithm will not be emphasized more than the other.
 
The performance that are being used in CC have nothing to do with each other, in terms of what they measure and how variable they are within each other. What you're suggesting is analogous to combining INT and FP scores together into one score. Does Spec.org do that? No. Why? Because they know it's wrong. The context of Spec.org's application of geometric mean is completely different from the context of CC.
 
The purpose of developing a fair scoring system is not to apply it to a single batch of past data and see how it would have worked. It is to make sure it gives a chance of winning to all contestants under a variety of cases. To directly show that geometric mean is not working for CC, I'll ask you one question. Look at MB (monkey_bollocks). They did poorly in Points and came last, because they're a small team, getting 15.1. Suppose, hypothetically, they did extremely well in both Conversion and Growth, and got the best scores in those, which are 97.7 and 250, respectively. Combine these three using geometric mean, you get 71.8. Now look at EVGA. Being the largest team, they got 120.3 in Points. Suppose, hypothetically, they did extremely poorly in both other categories, and got 52.2 and 79.1. Combine them using geometric mean, you get 79.2. You see what's happening? The smallest team cannot do better than the largest team even if they bust their butts and do best in two other categories, while the other does the worst. If you think this is fair, I have no words to say.
 
Why this is happening has nothing to do with geometric mean. It is about the fact that you're trying to use it in a very wrong context. What matters is the variability of scores within each. If in one category, the difference between the best and worst is larger than others, it's affect on the overall score will be greater. I don't know if you heard of ANOVA (Analysis of Variance), but this is a mathematical fact derived a long time ago. It is not my interpretation, it is a fact. 
 
If you really want to use geometric mean, first you have to set the ranges in each category equal (as in #671). Then, instead of summing the scores, you can use the geometric mean because then it will have a correct context of application (well if you care about balancing the category scores). I know for sure that your results won't change (comparing summation vs. geometric mean).
 
 

Hopefully I'm not dragging things in any direction - all I did was post data for people to examine and draw their own conclusions.

theGryphon

I'm gonna finish what I have to say by putting this final note on Stamina.

Stamina is a statistical model, and statistical modeling is not exact science. That means there is not one model that would do perfectly in all situations. How statistical modeling works is you develop a model, you use it; if it performs poorly, you tune it, or trash it completely. That said, Stamina, being such a statistical model, can perform very poorly, but that is not my expectation. The thing is, it is a very simple model, using only moving average. I could try and develop a more sophisticated and better model but it wouldn't be possible to explain it here. 

What's amazing is that you two guys, apparently with no experience, no know-how and no authority on statistics and statistical modeling, come and judge something you don't understand. By calling a statistical model, "meaningless", "trick" and "bogus", you're not insulting me, you're insulting the whole science of Statistics. Well done!

 
I certainly hope this is your final note on stamina.  To call what you have constructed a statistical model is a stretch at best.  Thanks for the lecture on statistical modeling but based on your claim of a graduate degree in management science I can be relatively certain that I have a much stronger background in statistics and modeling than you.  
 
When you say it wouldn't be possible to explain here is that because you are unable to explain it, or you don't think anyone here is capable of understanding it?  The fact that I don't flaunt my graduate degree on the same level as yours in a field that is much more statistically rigorous than "Management Science" does not mean that that I have "no experience no know-how and no authority on statistical modeling."  I suppose you might take me more seriously if I lead off with my credentials rather than the relevant topic at hand - because at this point that is the entirety of your argument.   
 
You can lecture all day about how great your statistics knowledge and credentials are, but what matters is results.  At the end of the day the primary driver of your "model" is random noise, rather than what you claim to be capturing. 
 
In regard to the geometric mean debate: gryphon's proposal is significantly different than geometric mean.  gryphon's idea, if implemented correctly, gives all categories a fair weighting where as geometric mean places a greater weight on categories with a larger spread.   

I withdraw my comment about stamina being "bogus" and apologize; I simply have not had time to read the stamina formulas in any detail, other than seeing the use of geometric mean and troy's comments.  "Artificial" would have been a better word, but I have to say that the emotional content of my words has been exaggerated ever since fact that you gave me a big thumbs down just because I didn't understand why you asked a particular question and so answered the wrong question; and then my dog died.  I don't want to further debate statistics knowledge or rebut earlier comments.
 
My intent in posting the spreadsheet was two-fold:
1.  Show that no matter what formula is used, the same teams place first and second.
2.  Show that no matter what formula is used, EVGA will not win.
 
Therefore, what is the point of further argument within our team?  It serves only to put lipstick on a pig.
 
If the issue with any of the formulas is that one category is different, there are a few ways to look at it:
1.  Fix that one category so it aligns with the others
2.  Remove the category (troy has been in favor of abolishing absolute points for a long time now)
3.  Acknowledge that the particular category is, in fact, the most important, and leave the fact that it might have a higher weight alone
4.  Force all the categories to fit into alignment
 
I think we have already covered 3 and 4.

It has become painful to read this thread.  Are we really this caught up about who wins the CC?  I thought it was about coming together as a team to promote folding, not battling over who has the best formula.  It seems like no one will ever agree.  How about we just aim to break our own personal/team records and let teams track whatever stats they wish.  I think having a winner in general has become the downfall of this challenge.  Maybe use your formulas to come up with a goal for all teams combined to reach.
 
I'm with this guy:

asus939 
so when does the CC start?

 

I'll throw in what AB would typically say here, with a little more spice.  Just because a person is being a horse's arse doesn't mean they represent the entire team, so don't base your actions on their/my behavior.  You may also know the mods here are very much hamstrung by corporate policies as far as what they can do; in a privately-owned forum, mods can be far quicker to bring the hammer down and not worry about legal threats.
 
P.S.  in the above, any reference to a horse's nether regions refers only to myself.

theGryphon

Punchy, I'm sorry about your dog. My sincere condolences...

Thank you all for the condolences.  Here he is, on the left, with his Basenji friend.  He came to us as an abandoned dog, obese and heartworm positive.  After two heartworm treatments that nearly killed him, he got healthy, dropping from 50 to 35 pounds.  My mother-in-law fell in love with him and took him back to California with her, where he was her loyal companion for years.  When she could no longer care for him, he came back to live with us again.  He lived through a stroke, multiple cancerous tumors, cataracts, chronic ear infections and resulting deafness, all without complaining.  We're going to send his ashes to be scattered with my mother-in-law's in the Sierras, as we know that will make them both happy.
 

I am still really thinking about joining in this time around.