Read this thread carefully. Use your intellect, which is not in doubt, to extrapolate a few things.

All points, except the early posts about my own machines, in this thread relate to Linux OS but the difference is a constant and points should be multiplied by 1.8181818 to give a direct comparison to WinXP results.

In the example given the Celeron 1.3Ghz machine finds itself in a quorum of faster machines(obviously). Due to the figures given early in the thread by Electyrolyte it can be seen that the faster machines are most likely of the Xeon 3.2 or FX55 speed range. The Celeron, has claimed 44 points and is awarded 24 in a Quorum of 20, 24, 44. (This is the correct result for the point system in use)

Likewise it should be obvious that if one has a computer or computers performing at the extreme (pun intended) end of the productivity scale, most, if not all, computers in your quorum will be slower and be therefore claiming more points than you. Your low claim is discarded by default and the highest also. One would be claiming 20 points per work unit but if the average is 36 or 37 that is what will be given, over a large sample of work.
Thus the points recieved for the fastest machines will be nearly double what they claim on average, likewise the slowest machines will receive only half what is claimed. There would be absolutely nothing to be gained by artificially inflating the "claimed" points of a "cutting edge" computer.

I would like to see a cut and paste of the quorum points results for some oif these fast systems. I feel certain that they will confirm this theory beyond doubt.

Cheers.

Then the subject is finished.
I will forward this to WCG and see if they think what you said is appropriate for a community advisor.
Have a good evening.

Perhaps we could put these questions on a more factual basis? First off, I notice this thread started off discussing points per hour, and somehow morphed into discussing points per work unit. The two are completely separate issues. High points per hour simply implies a fast machine. Higher points for the same work unit implies unbalanced benchmarks (intentional or otherwise).

I would have to call 27.85 points per hour terribly slow by my standards. 82.93 is a bit low, but not too bad. Until recently, my main cruncher was an AMD 4800+ overclocked to 2.64 GHz. It recently completed the following HDC WUs with these times and scores:

B04649_ 0120_ YTMA21-12-8-6-c1, 1.29 hours, score 18/18
B04643_ 0016_ CTMA4B-17-1-16-c1, 1.89 hours, score 27/32
B04640_ 0128_ CTMA1B-23-15-10-c2, 1.46 hours, score 21/19

This shows an average points per hour of 14.6 BOINC points, or 102.4 WCG points, and is representative of what that system routinely scores. I'm using standard BOINC 5.4.11.

And if you think that's bad, get ready for worse. I've just recently upgraded to a new E6400 Core 2 Duo overclocked to 3.5 GHz. It recently completed the following HDC WUs:

B04644_ 0035_ CTMA4B-17-10-24-c2, 1.47 hours, score 31/30
B04643_ 0026_ CTMA4B-17-10-1-c1, 1.24 hours, score 26/24
B04641_ 0159_ CTMA4A-17-21-14-c1, 1.33 hours, score 28/28

This gives an average of 20.3 BOINC points per hour, or 141.9 WCG points. Now that's a fast system!

And you're surprised that the team from one of the web's main overclocking sites has systems which average slower than a recent AMD X2 in power? The only reason the average team points per hour is so low is that they've scrounged so many older systems for their farms that it brings the average down! 87 points per hour is quite unexceptional for a bunch of AMD 3200+ or betters and/or fairly recent pre-Conroe Pentiums, especially from a bunch of overclocking fanatics.

As for the points per work unit, I still believe that the quorum system would inherently prevent any serious inflating of scores from even the most obscenely hacked client benchmarks. You can try to claim any score you like, but it'll just get thrown out. I've seen several of my result quorum members try to claim high scores, presumably using an inappropriate optimized client, and it doesn't do them any good. They still get the same, median score as everyone else.

Another data point. Just as a sanity check, I thought I'd calculate the average points per result turned in yesterday by the top 5 teams. I think we can all agree that the number of good results submitted in a day can't be hacked? And so, if there is any kind of rampant cheating going on, it should cause anomalously high points per result submitted in that day?

So here are the figures:

Easynews: 2,920,005 points / 9,702 results = 301 points per result
XtremeSystems: 948,569 / 2,918 = 325
Clubic: 739,665 / 2,132 = 347
Rochester Community & Technical College: 681,050 / 1,864 = 365
IBM: 493,652 / 1,530 = 323

I don't see any outliers here, unless maybe you want to go after those scoundrels at Rochester Tech. Ironically, XS is getting almost exactly the same points per result as IBM's own team.

So, just perhaps, the XS numbers come from having a lot of fast machines crunching through nearly 3K WUs per day?

Interesting. That's significantly different from the average figures.

Maybe XS's crackdown has paid off, and the averages are just a historical thing.

Unfortunately, since different projects have different work unit sizes, we're still left all at sea - if the points which are supposed to be our "level playing field" allowing us to ignore variation in computers and work units become meaningless, then we're left without a baseline.

That's why I prefer to use averages, since they smooth out some of the short term anomalies caused by work unit and project variation.