David Kim wrote on the Ralph board that he is extracting the data to remove some credits from people who have been using optimized clients backdated to February of this year(It would be further back if he had the data). How does he propose to do this accurately ? Will it be based on the machine type reported ?

This is a seriously dumb move which will cost the project many machines.

BTW, what is all this talk on Ralph of "Punishing" people ? Do you think we are naughty five year olds or something ? We are volunteers FFS.

---

G'day Whl

I think that the R@H team is attempting to stay true to the BOINC framework, lets face it, they owe BOINC for the computer power they have.

So they are thinking of trying to justly re-allocate the cobblestones/credits.

I have said before, that I think this is a bad choice for R@H as it will highly annoy many more BOINC'ers than if just allowing the new credit system to just come into existance.

BTW, what is all this talk on Ralph of "Punishing" people ? Do you think we are naughty five year olds or something ? We are volunteers FFS.

I don't believe that the R@H team would ever mean such a thing, they are quite aware of all of the contributions of their participants, which is why they are so active on the boards.

David Kim wrote on the Ralph board that he is extracting the data to remove some credits from people who have been using optimized clients backdated to February of this year(It would be further back if he had the data). How does he propose to do this accurately ? Will it be based on the machine type reported ?

It may prove too time consuming, but I believe it is possible to do this very accurately.

---

Hi Hugo,

I would like to know how this can be done accurately ?

I have an overclocked AMD64 @ over 2.6Ghz which was reported as a Sempron 3100+ for ages until I changed the OS. Sempron 3100+'s, if I remember correctly run @ either 1.8Ghz or 2.0Ghz depending on the model and socket,have much less cache and a lower HyperTransport speed etc. I'm sure many people will have innacurately reported machines like me as well.

Back later.

---

...
I would like to know how this can be done accurately ?
...

I guess they can do it by examining the WU type and the decoy count contained in the results, without even knowing what kind if machine/OS/clock speed you used to crunch them.

It will give most Linux machines a nice boost as those have been hurt most by the bad BOINC benchmark. So this is not all about reducing credits, it's about equalizing them.

Is there such a thing as a "WU type" though ? All WU's have different random seed numbers which take them off in all different directions even though they may have similar names. The decoy count varies wildly on my machines with similar named WU's BTW.

---

Is there such a thing as a "WU type" though ? All WU's have different random seed numbers which take them off in all different directions even though they may have similar names. The decoy count varies wildly on my machines with similar named WU's BTW.

It's being done per WU (not WU type), with a number of credits assigned per decoy produced, and this number of credits is determined from a sample run, so work units containing decoys that take longer to run will be assigned more credits per decoy. The only benchmark used is the machine's production, so in that way it seems like the credits are very well aligned with the production. The result will be that you'll be credited for the production of your machines - the CPU type/speed reported by the BIOS/OS isn't a consideration.

HTH
Danny

---

yes the decoy counts will jump around even with the same work unit (but different random seed). it should average out though.

From my understanding (feel free to correct if I'm off..):
FRA_t322_CASPR_hom001_6_t322_3_1zkiA_IGNORE_THE_REST_848_1079_90_0
FRA = The main Rosetta algorithm to be used on this WU.
t322_CASPR_hom001_6_t322_3 = casp t322 protein.
1zkiA = where we normally see the designation of the protein.
IGNORE_THE_REST = The secondary Rosetta algorithm to use on this WU.
848_1079_90_0 = most, or all of this is the random number.

Did you compare WUs that are identical except for the random number?

And (if you're not 100% AMD) are you comparing some P4 systems with hyperthreading turned on? Unlike an actual dual core cpu, the two threads don't always get an equal amount of CPU time.

(i.e. could you share the data from the WUs, and the makeup of the systems they ran on?)

---

From my understanding (feel free to correct if I'm off..):
FRA_t322_CASPR_hom001_6_t322_3_1zkiA_IGNORE_THE_REST_848_1079_90_0
FRA = The main Rosetta algorithm to be used on this WU.
t322_CASPR_hom001_6_t322_3 = casp t322 protein.
1zkiA = where we normally see the designation of the protein.
IGNORE_THE_REST = The secondary Rosetta algorithm to use on this WU.
848_1079_90_0 = most, or all of this is the random number.

Did you compare WUs that are identical except for the random number?

And (if you're not 100% AMD) are you comparing some P4 systems with hyperthreading turned on? Unlike an actual dual core cpu, the two threads don't always get an equal amount of CPU time.

(i.e. could you share the data from the WUs, and the makeup of the systems they ran on?)

At first I thought that for this system it doens't matter what computer setups the tests are run on - if a decoy takes 30 mins on one machine, and then takes 40 mins on another machine, its the comparison of these two run times that's important. It's fair to expect the first computer to consistently take approximately 3/4 of the time of the 2nd machine, and therefore claim 1/3rd more credits.

I assume what you're getting at BennyRop, is that in testing, the median may be shifted by the makeup of computers as (for e.g.) a P4 may suffer on an FPU intensive decoy, while an Athlon will fly through it. If there are 10 Athlons and 1 P4 used in testing, then the decoys will be run through quickly, and be given a low score. The P4 will then take a long time to battle through the decoy, and pull the median for the decoy up slightly. If more P4s were used in testing then the median (nd therefore the credits assigned) would be higher.

Obviously, if decoys from all WUs are affected similarly in the P4/Athlon test then there is no problem and no need to test on different setups. If not, and the P4s fare better on other WUs then this could shift the median and affect the credits assigned slightly.

I think the outcome will be pretty minimal, but there will probably be some weighting by configuration, so at least initial testing needs to be done on a range of computers.

---

David Kim wrote on the Ralph board that he is extracting the data to remove some credits from people who have been using optimized clients backdated to February of this year(It would be further back if he had the data).

Just to clarify, Kim never said that they plan to impose the fair credit system on older results. They were responding to a request to see if it was possible. They won't actually do it unless there is a strong consensus among crunchers that it should be done. I don't see such a consensus, so I expect they won't try to fairify the past.

David Kim wrote on the Ralph board that he is extracting the data to remove some credits from people who have been using optimized clients backdated to February of this year(It would be further back if he had the data).

Just to clarify, Kim never said that they plan to impose the fair credit system on older results. They were responding to a request to see if it was possible. They won't actually do it unless there is a strong consensus among crunchers that it should be done. I don't see such a consensus, so I expect they won't try to fairify the past.

To clarify further - it's not to 'remove credits from those running optimised clients' - it's a full assessment of all the credit due for all Work Units run, to even out all the reasons for variations, including the client used, the hardware (P4s seem to have been hard-hit on the current system), the OS, and probably the amount of cache and RAM available (which might well not have an effect on the benchmarks but probably will on the crunching ability).

HTH
Danny

---

David Kim wrote on the Ralph board that he is extracting the data to remove some credits from people who have been using optimized clients backdated to February of this year(It would be further back if he had the data).

Just to clarify, Kim never said that they plan to impose the fair credit system on older results. They were responding to a request to see if it was possible. They won't actually do it unless there is a strong consensus among crunchers that it should be done. I don't see such a consensus, so I expect they won't try to fairify the past.

(RALPH is down now, so I can't provide the exact quote and URL.)

They said the new system, once approved and official, would be applied to all credits retroactively since February. No polling for consensus of the crunchers. Decision made.

All of us who have been working since February with a set of rules have been played.

---

Reno, NV
Team: SETI.USA

David Kim wrote on the Ralph board that he is extracting the data to remove some credits from people who have been using optimized clients backdated to February of this year(It would be further back if he had the data).

Just to clarify, Kim never said that they plan to impose the fair credit system on older results. They were responding to a request to see if it was possible. They won't actually do it unless there is a strong consensus among crunchers that it should be done. I don't see such a consensus, so I expect they won't try to fairify the past.

(RALPH is down now, so I can't provide the exact quote and URL.)

They said the new system, once approved and official, would be applied to all credits retroactively since February. No polling for consensus of the crunchers. Decision made.

All of us who have been working since February with a set of rules have been played.

One can only be "played" if one allows it to happen.
Need I say more? Lets wait and see what is going to happen.
Movieman

---

One can only be "played" if one allows it to happen.
Need I say more? Lets wait and see what is going to happen.
Movieman

Are you saying it is our own fault for not being able to see into the future, and know that they were going to change the rules retroactively? Seriously?

R@H lured people with false promises. Period. Their rules were written in plain english. R@H time after time delared that optomized clients were *not* against the rules, in this very forum. People joined and operated under the rules in good faith. R@H has turned themselves into liars.

They want to change the system going forward? Fine by me. Just don't change it going backwards.

Had people known, they may have chosen to not join, or go elsewhere. Now they have lost out on those other potential credits. Will R@H be making up for that lost opportunity? Not a chance.

And having retroactively changed the rules once, they will probably do it again in the future. Not a good way to get new members to joing with a rep like that.

No need to "wait and see". They already started in black and white that they will be recalculating the scores since February. It's a done deal unless they change their decision.

---

Reno, NV
Team: SETI.USA

They said the new system, once approved and official, would be applied to all credits retroactively since February. No polling for consensus of the crunchers. Decision made.

I think David said they will be showing both the old credits untouched and the new system alongside (2 columns). The way I read it, it wasn't a done deal and they wanted to see what the community thought first by showing both systems next to each other.

One can only be "played" if one allows it to happen.
Need I say more? Lets wait and see what is going to happen.
Movieman

Are you saying it is our own fault for not being able to see into the future, and know that they were going to change the rules retroactively? Seriously?

R@H lured people with false promises. Period. Their rules were written in plain english. R@H time after time delared that optomized clients were *not* against the rules, in this very forum. People joined and operated under the rules in good faith. R@H has turned themselves into liars.

They want to change the system going forward? Fine by me. Just don't change it going backwards.

Had people known, they may have chosen to not join, or go elsewhere. Now they have lost out on those other potential credits. Will R@H be making up for that lost opportunity? Not a chance.

And having retroactively changed the rules once, they will probably do it again in the future. Not a good way to get new members to joing with a rep like that.

No need to "wait and see". They already started in black and white that they will be recalculating the scores since February. It's a done deal unless they change their decision.

I think Movieman is of the same opinion as you (which highlights how easy it is to misinterpret on internet forums!) As Sangamon says, a system has been devised that can allow the comparison of the claimed credits and what they should be if they related to the work done. They're backdating the new system and then the two systems can be compared. We're being given more info- not less.

Danny

---

I think Movieman is of the same opinion as you (which highlights how easy it is to misinterpret on internet forums!) As Sangamon says, a system has been devised that can allow the comparison of the claimed credits and what they should be if they related to the work done. They're backdating the new system and then the two systems can be compared. We're being given more info- not less.

Danny

Over on RALPH, they said very clearly that the stats since february would be converted, once we move to the new scoring system. Am I the only one who read it that way? Perhaps I misread it, but I don't think so.


I would show you the post, but RALPH is still down...

---

Reno, NV
Team: SETI.USA

David said he would look into estimating 'work' credits based on 'boinc' credits from Feb. So the 'boinc' credits would be untouched, and the 'work' credits would be consistent backwards and forwards. But I didn't get the impression that anything was 100% decided yet.

It's also clear that there are still quite a few bugs to be sorted out, and a lot more work to do, before things settle down.

-- Edited for clarity

---

my impression was that it would be possible to recalculate stuff till february after somebody asked if it was possible, it would be more effort than it was worth though most likeley, and i did not get the impression that anything in that matter was decided yet.
just to give you one more possible interpretation :P

(i can not say that i saw the last posts on the ralph forums though, not sure when it went down)

One can only be "played" if one allows it to happen.
Need I say more? Lets wait and see what is going to happen.
Movieman

Are you saying it is our own fault for not being able to see into the future, and know that they were going to change the rules retroactively? Seriously?

R@H lured people with false promises. Period. Their rules were written in plain english. R@H time after time delared that optomized clients were *not* against the rules, in this very forum. People joined and operated under the rules in good faith. R@H has turned themselves into liars.

They want to change the system going forward? Fine by me. Just don't change it going backwards.

Had people known, they may have chosen to not join, or go elsewhere. Now they have lost out on those other potential credits. Will R@H be making up for that lost opportunity? Not a chance.

And having retroactively changed the rules once, they will probably do it again in the future. Not a good way to get new members to joing with a rep like that.

No need to "wait and see". They already started in black and white that they will be recalculating the scores since February. It's a done deal unless they change their decision.

I'm sorry, you misunderstood me. You and I feel VERY closely on this issue. We are not at odds on it at all.
What I was trying to say(badly it seems) is that one can be played yes, but only once. Any more than once and your letting it happen ..IE: The old line:
" Fool me once shame on you, fool me twice, shame on me."
Your last line is exactly correct, it's a done deal unless they change their decisions.
That is exactly what I am waiting to see. With age(54) I've learned a little patience and not to expect perfection from anyone.
Thanks for your time,
Movieman

---