| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: BOINC Points post in Member News |
| No member browsing this thread |
|
Thread Status: Locked Total posts in this thread: 277
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
THe D930 @3600 is an outlier in that it consistantly claims 44-48, very low. I want to know for fact if that is pulling the quorum down as it appears to often. There have been, so far, a couple of times it scored significantly higher upto 78 so was obviously excluded as a low score but in others it scored a lot lower appearing to pull the quorum down.
----------------------------------------The 78 was a lot higher than the average was this due to a non stock Boinc being in the quorum ? If so then this won`t happen as often and the D930 score will drop. If we average the points over the three machines and 30 units it works out at 62 per faah so the 78 was an anomoly and WAS NOT longer in time than others. I will await until I have more work to give better judgement. Anyone else doing a similar comparison would be welcome ! [Edit 1 times, last edit by Former Member at Nov 15, 2006 11:13:11 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
{snip}If u had a quorum of 45, 48, 58, one gets e.g. 50, which is 2 above what would be given under the old rule. In effect, 6 more credits in total are handed out over the old system. {snip} cheers Hi all I'm back after some rest...  Sek: The top score is eliminated from ALL quorum calculations and this quorum would give a score of 46.5, not 50. knreed's proposal looks good to me, pity about the non poductive clocks, scientifically speaking. Doesn't seem a lot of point trying to pull it apart. The submission is in and that's that I guess. Must agree though that an averaging mechanism seems unnecessary. I think most would agree that somewhere around 65 seems right. If one were to take into account the speed and type of computer the 'cobblestone' system was designed for, it would probably be well over 100 points. Carlh is right, the faster the average computer gets the lower the average score will become. This is why older equipment has suddenly started to throw 'outliers' into the quorums. The recent influx of faster equipment has pulled the quorum averages down significantly and the removal, in most cases, of optimised client usage has now given a truer perspective to the situation.  The new system should addrtess that situation, if and when it arrives.Ever onward. Keep crunching Cheers. ozylynx |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I hate to contradict you, but the new credit system no longer automatically discards the top and bottom claims. Instead, the deviation is calculated, and outliers beyond a certain theshold are discarded, and the remaining claims (usually all of them) are averaged. This theshold is not the same as the penalty theshold - that is at least twice the threshold at which claims are ignored. Only extreme outliers are penalised.
So, you see, it is not a simple mechanism, but it strives to be fair. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
No sir Ozylynx.... the 58 is not an outlier, therefor adds in full for the quorum average!
----------------------------------------
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Nov 15, 2006 2:18:20 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I stand corrected gentlemen.
I hadn't seen anything about this till now. I've been watching my slowest machines of late, for obvious reasons, now I see that they will be effectively penalised even on the rare occasions that they aren't excluded outliers! On looking more closely a 1Ghz machine, a quite acceptable to fast computer not all that long ago, seems to have about a 50/50 chance of being excluded from calculations. Slower than that and you're in real trouble.!! Is there also a mechanism in place to find low score outliers? Hopefully this system will be able to be removed when the new benchmarking arrives. It's starting to bite Joe(slightly less than)Average. Here's hoping. Cheers. ozylynx |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Instead, the deviation is calculated, and outliers beyond a certain theshold are discarded, and the remaining claims (usually all of them) are averaged. This theshold is not the same as the penalty theshold - that is at least twice the threshold at which claims are ignored. That being said then, in this instance, the D930 with the low claim would ALWAYS be included therefore lowering the score of the whole quorum ? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Here's the other end of carl.h's situation, also pulling average quorum scores down.
faah0892_ bdb219_ mx7upj_ 06 Valid 11/11/2006 21:01:48 11/12/2006 10:31:19 8.27 50 / 53 faah0892_ bdb219_ mx7upj_ 06 Valid 11/11/2006 20:58:36 11/12/2006 08:16:13 8.52 55 / 53 faah0892_ bdb219_ mx7upj_ 06 Valid 11/11/2006 20:47:40 11/12/2006 19:53:08 18.75 96 / 53 This is obviously not an optimized client situation. It is an older computer just struggling to keep up. Over 10 hours longer than the others to complete the task.  Would it be possible to include a time taken check in the calcs to determine outliers? Cheers. ozylynx |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
I have pulled the data from 166 quorums and here are the results (I added additional logging to record the old value and the new value so these were the first 166 sets validated since I added the logging).
In 71 cases the new method resulted in a lower credit granted. The average amount lower was 1.9 credits. In 95 cases the new method resulted in more credit granted. The average amount more granted was 2.5 credits. (none were identified as equal becuase these were double values and they didn't match exactly) Over the full set of 166 quorums the average amount granted under the new system was 0.61 credits more then under the old system. |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges:
|
In response to a question:
----------------------------------------Yes both high outliers and low outliers are removed from computing the credit granted. Only 'extreme high outliers' are penalized. [Edit 1 times, last edit by knreed at Nov 15, 2006 4:27:58 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
 I am pleased to hear this. I know my machines are fairly consistent, but in ways that differ from what knreed say is 'normal'. Looking at the quorums in my results status page, I see some computers that seem horribly different. I always wonder if some posts are not being made by people with one of those machines. If you continue to gather statistics, it would ?possibly? be helpful to post something with a breakdown more detailed than we see on the By Project statistics page. This isn't urgent as it will take a while to work out the details for a new BOINC points system, but it would be helpful in gaining informed commentary. (Or just comments, informed and otherwise.  )Lawrence |
||
|
|
|