Folding Forum

k1wi wrote:

orion wrote:

vbironchef wrote:Exactly! All that is happening is that people are turning F@H off. I don't think it is very hard to understand, do you?

In the 8 years my team has been folding it has had 368 folders but only 25 active folders now.

People come and go for all sort of reasons...it's the way of life and the way of folding.

The fact that people come and go does not mean that we should not explore how to encourage people to stay for longer.

But while trying to find ways to keep a certain group of people folding, which is commendable, you’re also alienating some that may stop folding.

It's a two edge sword.

not reward them over time for the improvements that Intel or GROMACs make!

Yes I understand.

However, if everyone took that view, we would all be using P1 computers.

I have a single core laptop(well my P1 266GHZ does still work but it is so slow I can't stand it anymore) that runs XP just fine.
It has been "good enough" for almost 9 years now.

The multiplying improvements of INTEL, et. al. is exactly why progress can be made on this project.
Not recognizing or rewarding that progress seems counter-intuitive to me.

Mdk I'm with you, I like seeing the millions ppd some are making

But in 5 years, what will they be making? Can you imagine needing a dual monitor setup to read EOC stats because the credit and ppd fields have become so wide the entire content can't fit on one screen.

I initially said: what's wrong with factorizing ppd there? That would prevent the display problem.

I still wouldn't mind that at all, as it would make it easier to compare scientific contributions.

But, the only valid point I agreed with was the exponential increase in computational power. So a discussion about how to 'fix' that problem is maybe not relevant now, it would get more relevant over time. Nothing wrong with starting a discussion early.

How would you keep the system intact and deal with the numbers we will be seeing in near future?

@vbironchef

I have cards of that age Folding@Home never been focussed only on the top end.

mdk - under my proposal people would not still be folding on p1's, because p1's would still be making the same proportional # of points as they do under the current system, it's just the absolute values aren't increasing exponentially.

It's time to make new suggestions, as that's what's going to make you the one who came up with a workable solution.

Can you agree with me MtM that what we have come up with together, along with all the other imput is:

Revised PPDn = PPDn / Yn (also written as "Revised PPD = PPD / Y"), where Y is determined by PG based on whatever methodology that they think suits, and where changes in Y is publicised - so that from time n to time n+1 etc people an understand why their current PPD has been normalised to the level it has.

This then makes:

Total Points = sum(PPD1/Y1 + PPD2/Y2 + ... +PPDn/Yn).

However, I don't think that PG can use PPDn/Yn on it's own. There needs to be a way of incorportating PPDn / Yn into the following formula:

ppd = base_ppd * speed_ratio * max(1,sqrt(x*speed_ratio))

However the following:

ppd = base_ppd * speed_ratio * max(1,sqrt(x*speed_ratio)) / Yn

does not solve the effect and this can be seen on the following graph, where over time the purple line is still increasing:



I put this down to the fact that this is because 'technological improvement' is having a compounding effect. Once in increasing the baseline PPD and then again in increasing the speed ratio.

Alternatively, and now I think about it, the following formula needs to be adjusted so that the concept of PPDn / Yn remains constant over time:

final_points = base_points * max(1,sqrt(k*deadline_length/elapsed_time))

Over time the value 'final points' fora given WU should actually decrease, because every 18 months the rate at which WUs are completed doubles and this should cancel it out.

I wonder, therefore, whether the following holds that true?

final_points = base_points * max(1,sqrt(k*deadline_length/elapsed_time/ Yn)) / Yn

K1wi,

As I mentioned, it comes down to your core philosophical beliefs.

I have subscribe to the black and white view that symbols should have a direct correlation to what they represent.
Your system continually "re balances" that correlation to fit a predetermined outcome.
This goes against my world view. I want to see facts lead to outcomes. Not facts redefined to fit outcomes.

I not saying this doesn't happen in real life. We have fiat currency right?

I just don't like it.

mdk777 - I don't agree with "re balances" that correlation to fit a predetermined outcome.

But I agree my core philosphical belief is that it is better to avoid the situation we have now where points increase exponentially over time because technology is improving exponentially.

Why do I want to avoid that situation? Because eventually, if FAH continues, then we will be measuring ppd in factors and the factors will also double every 18 months and then something will have to be done, because even the factors will get too large. I identify the cause of this to be underlying technological improvement, therefore I direct a proposal, with the feedback of other users that creates a community response, that solves this underlying issue.

If it is not the proposal debated in this thread then it will have to be something as difficult as dropping the 0's off everyone's PPD and possibly also wipe the same number of 0's off user's total. That would effectively make a lot of people's contributions over the years dissapear. I don't want to do that!

I appreciate that you don't agree with my proposal, but at this stage it is only that, a proposal being formulated by the community. If you don't agree with it and it becomes a final product that can be considered for replacing the existing model then by all means, argue then why it shouldn't be implemented. At the present you have outlined your reasons why it won't work and those reasons are something that people working on the formula will have to take into account.

I initially said: what's wrong with factorizing ppd there? That would prevent the display problem.

I still wouldn't mind that at all, as it would make it easier to compare scientific contributions.

But, the only valid point I agreed with was the exponential increase in computational power. So a discussion about how to 'fix' that problem is maybe not relevant now, it would get more relevant over time. Nothing wrong with starting a discussion early.

How would you keep the system intact and deal with the numbers we will be seeing in near future?

Yeah 10 to the 99th power gives you a fair amount of headroom as far as I can see. and 10 to the 999th power a bit more.
I don't see how math notation is going to be overwhelmed by this problem. I really don't see computational power continuing along Moores curve indefinitely...and if it does, problem solved

all the possible computations will be done and the project ended.

But as I said, we are looking at a specific solution without first debating what we really want to achieve.

The problem I have found with people debating what we want to achieve is that those people complain that there is no specific solution for any solution to achievement...

Why do I want to avoid that situation? Because eventually, if FAH continues, then we will be measuring ppd in factors and the factors will also double every 18 months and then something will have to be done, because even the factors will get too large.

Yes, I reject your premise.

But I have made my view clear I hope.

I will move on and let others continue the discussion.

I respect that mdk777, but in that case let this thread play out. Continuously stating that the premise is wrong isn't helping create an end formula that can then be debated as to whether it achieves whatever the 'community' eventually decides it wants to achieve.

mdk777 wrote:Re: Suggested Change to the PPD System

Post by mdk777 » Mon Mar 19, 2012 4:53 pm

Why do I want to avoid that situation? Because eventually, if FAH continues, then we will be measuring ppd in factors and the factors will also double every 18 months and then something will have to be done, because even the factors will get too large.



Yes, I reject your premise.

But I have made my view clear I hope.

Good, because mdk777 just doesn't get it, nor does he want too.

If you think folders have a unlimited budget or want to buy the latest and greatest computers for F@H, your wrong, really wrong. Sad. The whole point of this thread was on how to make the PPD system fair. k1wi don't give up! You are on the right tract. mdk777 is just not using any common sense.

I don't think he is not using common sense, I just accept that he has a different perspective on the points system.

There will be a time at a later date where the different perspectives need to be considered, but there is merit in people stating during the decision making process here that there are different processes.

He has done that and it can be noted and I think it can improve the decision making, but most importantly, improve the explanation, because at the end of the day, the devil is in the explaining.

He has also said that now that he has stated his position he will let the thread develop and I appreciate and thank his stance.

@mdk777

Well what we want to achieve is preventing the need to display the ppd/total credit using a factor. Nothing more nothing less, as far as I'm concerned. I don't know how to fix that without normalizing the ppd, which in effect would normalize the totals ( not entirely, as totals will eventually still grow very big but not as soon as when not fixing the exponential growth in ppd based only on computational increments ).

@vbironchef

mdk777 is not sad, nor is he not using common sense. He wants, just as I do, to be able to correlate ppd to science. That's an historical premise of f@h, donor's can see how much they helped the science but looking at their ppd. You can not claim that is sad, as the other way to measure points is always based on keeping as many donors happy and not on the science.

I concur with the reason he doesn't want to normalize, I just also know it will need to happen at some point in the future.

@K1w1

As to describing the solution as a common proposal, even that's going to far as it's something which was mentioned already a really long time ago, and as you pointed out the analogy with a monetary system it's also something which has come up for many different problems. It's an universal solution to an universal problem, with the added complexity that Y in a monetary system is already defined. I'll be more then happy to call this the k1w1 formula if you succeed in making it work It's to early now to be claiming 'ownership'. Sorry btw if I was harsh, that's not the intent. It's safe to say I should have been less negative at the beginning of this thread as you're not just repeating things which were brought up before. I appreciate your efforts into making this a point of discussion, it's something which even if people don't see it now, can not be avoided as a point of discussion at some point in time.

I also looked at the qrb formula and failed to put Y in there in a way which would work As I hinted to already, I'm quite stuck here besides saying: ppd = ppd / y and let PG come up with Y by benchmarking common fahcore functions ( which I proposed as well as keeping historical work units and rebenching them each time ).

Your graphs show that QRB will still make the ppd growth exponential, and I agree. But I can't agree with changing it to be less then exponential unless you can come up with a way to prevent people running multiple instances instead of one. PPD is not only meant to measure, it's also meant to give directions on how to tun the project. Even forcing each client to always use all resources for one instance would just make people use vm's to run multiple instances unless you give an incentive to return work unit's quicker.

Btw I have to admit I don't like the graph layout, time should go from left to right, and I don't understand the flat sections before they start raising. I'm also color blind to the point where your legend isn't clear so I'm having to guess which line is which revision ( though that I think I'm not making a mistake on ).

What we want is a fixed line PPD, and as a result, without looking at the formula, the QRB should be a fixed line above the normal base PPD. But as I said already, a fixed line equates to a linear increase, and that I don't think is something which makes the cut. PPD as a fixed line would need QRB as incremental line above it, but resetting it on each normalization. That would make the QRB line still grow, but it won't grow exponentially over time. The manner to reset it, would be using Y in relation to speedup factor. Edit ( just as you proposed already in the post with the graphs ). What happens in the graph if you do that? ( sorry, 1:23am don't feel up for making my own graphs, been a long and exhausting day and I'm going to bed ).

Let's take a new video game for example. A new player doesn't expect to be on level 50 from the beginning, right? Same goes for F@H. you have to put your time in. Even if you have the latest and greatest rig you start at level 1. So you play the game for a couple of months and a new player shows up with a faster rig than yours, so what happens? He chops you into little pieces! Did he put in the time and effort you did? Of course not! Are you upset? Answer, yes! Do you go out and buy a faster computer or do you stop playing the game because it's not fair. I choose not to play anymore. Simple as that. What's the point, there will always be something new and improved that costs to much. Do you really want to continue to play, I think not. Same applies to F@H. People should have to level up and pay their dues before they can slay the dragon.