Thank you for that detailed and honest response.

The quorum system is good, it discourages out-and-out cheating, and lessens the effect when it does happen. But that's not the reason for the quorum. As I explained in another thread, the quorum is needed purely for scientific reasons.

BOINC points have always been a mess. It's a difficult subject. Should they reward effort? Should they reward science? What we do know is that since all the work units are a different size, a really simplistic solution wouldn't be fair.

Sadly, we have to work with what we've got.

Using an optimised BOINC client with an optimised science application is perfectly fair - more work, more points. Running more than one project can lead to this inflated benchmark being applied to projects that can't take advantage of it. The attitude around here is: "we don't care how expensive your computer is, you should get points according to the work you actually perform". I know it would be better if the science could be optimised further, but it's simply not practical. The reasons for this are again explored in depth in the other thread.

So, in the interests of fairness, I think anyone dissatisfied with the standard client should use a calibrating client. Truxoft have deprecated their optimised client and recommend their calibrating client. This lets you claim different benchmarks for different projects, so you can keep up with the SETI superusers, and not outrage non-optimised projects with impossible claims. However, the calibrating client should award you slightly more if your system is well configured and can perform well for a particular project. Not an impossibly inflated score, but a fair score.

Taking all this into account, I would feel a lot happier if the official stance from your team leaders (and any other team with killer point aspirations) is to only use a calibrated client, at least until that happy day when we get a genuinely fair point system.

And the final word: WCG are aiming for the average user. The man in the street. For every super-cruncher, WCG has thousands of perfectly unexciting crunchers to replace them. So, while we love having you here, and healthy competition will hopefully raise us to even greater heights, the WCG admins can and will act swiftly if things get ugly.

Steady on there, I'm only a mere Team Captain and Community Advisor. I may have the inside track, but I'm a volunteer and I can't speak for World Community Grid. My opinions are my own.

I apologise if you found me confrontational. It's just that I've noticed a tiny minority of your team mates pretty much flaunting the fact that they're prepared to claim every point they can, no matter how fair it is. That causes friction, and isn't good for the project as a whole.

I was merely trying to preempt any of that. Better to keep things out in the open and discuss the facts without letting personal issues cloud the debate.

In the past, WCG have been so successful by avoiding ticking any group off too badly. As no doubt you realise, doing exactly what everyone wants out of a crunching project is simply impossible. So, we're searching for a compromise.

The WCG techs already have improved the scheduler in various ways. For example, "late" work units are reissued to a fast computer so that validation isn't delayed. If you can demonstrate that a faster CPU has a valid reason for claiming more points for a work unit than a slower one, then I'm sure the WCG techs will consider a feature as you suggest. However, the theory is benchmark x CPU time will be equal on any computer for the same work unit, if the benchmark is good enough. The variation in claimed credits we see demonstrate that it isn't good enough, but it's the best we have.

Maybe we should persuade the BOINC guys to add the calibrating stuff to the official client. Then all the point claims will be similar, and the quorum won't affect points so much.

You're welcome.


When needed for scientific reasons it is a great thing to have, but I must disagree that it stops or lessens cheating. It is not that it doesn't, but rather that it does not stop it as effectively as other methods could. A lack of effectively stopping something only shows weakness.


They should reward both on a 1:1 ratio. One unit of work done gets one unit of reward. Given work unit size or credit recieved this could approach a 1:10 or more ratio, but the important part is to keep the ratio equal at all times. Never reward more for one item than another, and as such never give out more work for less reward. An analogy would be me and you shoveling a sidewalk. Maybe I get $5, you got $2.50. Assuming we performed the same amount of work over the same time period that would be wrong.

However, if we had differing work loads maybe rewarding each shovel full with .05c would work. Even if we are using different sized shovels the one with the larger shovel should get done faster and go onto new work faster. Its a self-regulated system.


Yes :(


I dont think any thread has shown the performance difference between the different subsets of instructions. While the "we don't care how expensive yoru computer is, you should get points according to teh work you actually perform" quote is nice, the flaws in applying reward to work performed show the opposite. This is the crux of the issue and sadly it seems distributed computing projects organizers have a hard time realizing it. This shows weakness.


Again, the standard system in place is flawed. This means all users are being short changed (or could be getting over-rewarded). Moving to a different client does not fix the issue for everyone else. This means the calibrating client is also flawed as it relies on the stantard system to create its baseline.


No, that only tells the project leaders we accept that they do not recognize the work we are performing. Rewarding equal to work performed shows you value the work you are being donated.


It should never turn ugly in the first place. If it is allowed to turn ugly, it shows weakness.

This is what worries me and kept me from joining WCG from the start with the rest of my team mates. By far and large most credit/reward systems for most DC projects are flawed. This creates tension among the user base and eventually us stats junkies who use stats as a reason to put more and more money into these projects just leave because of getting bad mouthed. We have high proficiency because we perform well per unit of time. Stats give us a reason to continue trying to do the best for a project we can. Taking away our drive or not rewarding equal to work performed only makes it harder for us to justify donating our money to you. That is after all what we are doing.

Well. I am a member of the XS team and i am not running the optimized client. But that doesnt make me a 2nd class member.

Everything you write can be shortened. "i dont want that optimized clients are used." - You may add some justifications for this, but i suggest not to insult us. Really. Your post has a way uglyer badwords/goodwords ratio than southpark the movie.

To the XS Team:
Hey, I really like this place. Lots of nice people. Friendly. Good admins.
Great medical app that is well worth our efforts.
Let's do what we can to help this app.
Have a nice day guys!