One last request for backdating of credits.
Message boards : Number crunching : One last request for backdating of credits.
| Author | Message |
|---|---|
John McLeod VII Send message Joined: 17 Sep 05 Posts: 108 Credit: 195,137 RAC: 0 |
Please hear me out before deleting the post. There is one reason that was never brought up. There are two different sets of rules that people were working under. Those of this project which are very liberal about fluffing credit scores, and other projects where all the credit for someone who is caught fluffing their credit above the average were deleted. Therefore there were at least two sets of rules in place. One set of rules for those that crunch only for this BOINC project, and a different set of rules for those that crunch for many BOINC projects. I am creating this thread because I can no longer find the original discussion of this.   BOINC WIKI |
 carl.h Send message Joined: 28 Dec 05 Posts: 555 Credit: 183,449 RAC: 0 |
John, I understand your point completely and have stated so elsewhere. What is a worry at the moment is the new systems parity to Boinc seeing as it is not, as I know, a boinc measure. I think this is definately worth a question from the cross project Boincers. I`m no expert in the Boinc field and there maybe other projects that don`t give parity through Boinc, I just don`t know. Would my machine earn the same amount of points across all the Boinc projects ? Not all Czech`s bounce but I`d like to try with Barbar ;-) Make no mistake This IS the TEDDIES TEAM. |
[B^S] thierry@home Send message Joined: 17 Sep 05 Posts: 182 Credit: 281,902 RAC: 0 |
A post from Bruce Allen (Einstein@home leader) on the E@H board : My intention is a simple one: ON THE AVERAGE a host machine running Einstein@Home should get the same number of credits/cpu-hour as a host machine running the other BOINC projects that grant credit. Here ON THE AVERAGE means averaged across all the hosts that are attached to multiple projects, and averaged across all the projects (suitably weighed by the number of cross-project hosts). Rationale: this way, people will chose projects based on their scientific and other merits, and likelihood of success and impact, NOT for other reasons such as credit granted. Corollary: assuming that other BOINC projects do the same, this will tend to make hosts move to the projects that they are best suited for. Cheers, Bruce |
|
carl.h Send message Joined: 28 Dec 05 Posts: 555 Credit: 183,449 RAC: 0 |
Thierry does a system of parity across Boinc exist ? I take it not from the Bruce post, that being the case Boinc totals are pretty much meaningless imho. Not all Czech`s bounce but I`d like to try with Barbar ;-) Make no mistake This IS the TEDDIES TEAM. |
|
tralala Send message Joined: 8 Apr 06 Posts: 376 Credit: 581,806 RAC: 0 |
There are two different issues: a) backdating credits b) making sure the new credit system obeys boinc project parity As for b) the post from thierry sums it up very nicely and is indeed which we should aim for as well. I'm curious what Einstein will do, since at the moment they are the project which grants the most credits/hour. As for a) this is a really difficult situation. As John stated correctly people operated under two different sets of rules. There are valid reasons to backdate as well as to not backdate. Furthermore this question considers fairness and justice as well as trust, reliability, consistency etc. from very different angles and there are two different viewpoints which both are valid. So I don't think one can come to a just decision. Backdating is seen by some as the only just way while others disagree. I don't think we will ever find a consensus on this and think under these circumstances it is the correct decision to leave the past behind and concentrate on making a fair system for the future. |
|
John McLeod VII Send message Joined: 17 Sep 05 Posts: 108 Credit: 195,137 RAC: 0 |
Thierry does a system of parity across Boinc exist ? I take it not from the Bruce post, that being the case Boinc totals are pretty much meaningless imho. Most of the projects are fairly close, and there are discussions among the project leads of several projects to attempt to get the various credit granting methods to have the same rate for a given host. It cannot be perfect because some CPU types are optimized for different operations, and some projects have lower or higher locality than others so L1, L2 sizes and RAM speed all matter more than expected. BOINC WIKI |
Fuzzy Hollynoodles Send message Joined: 7 Oct 05 Posts: 234 Credit: 15,020 RAC: 0 |
... As a crossproject cruncher I can say yes, in principle. And Eric Korpela at Seti is working on this, as I've understood it. Does anyone have more information on this? [b]"I'm trying to maintain a shred of dignity in this world." - Me[/b] 
|
|
[B^S] thierry@home Send message Joined: 17 Sep 05 Posts: 182 Credit: 281,902 RAC: 0 |
I suppose it was the hope of D. Anderson. But ... this is theory, a good idea but theory. You knon what can happen if you give the same tool or machine or idea or ... to ten guys and say 'Do xxx with it'. After a certain time .... Edit: this was an answer for Carl (I need to type faster ;-)) |
|
[B^S] thierry@home Send message Joined: 17 Sep 05 Posts: 182 Credit: 281,902 RAC: 0 |
There are two different issues: I used the E@h optimized app (v4.24) when it was in test and get 176 credits per WU for +/- 8h30' of work. Then the credit decrease to 148 when the v4.24 became live, and we are now at 122 credits (still for 8h30'). |
Saenger Send message Joined: 19 Sep 05 Posts: 271 Credit: 824,883 RAC: 0 |
There are two different issues: I agree. The disparity is not only between Rosetta and the rest, but also within Rosetta. Most people will use the stock client, as they don't read all this here and just want to join the science or whatever. Some use the overclaiming clients. Both got granted and the stats don't reflect the highly uneven playing field. To get e real overview who did what in regard of scientific input (i.e. more calculations), a backwards correction would be fine. Whether the outcome would be to triple the credits of stock client crunchers, or to delete credits from the overclaiming crunchers is from the inner-project POV irrelevant, not of course from the BOINC POV. I think it would have been best to keep the cat in the bag, so nobody would have asked for it, as it's imho quite unusual to have such a long backlog of detailed individual result data. But now the possibility is common knowledge, and imho should be done to get real stats. |
|
Fuzzy Hollynoodles Send message Joined: 7 Oct 05 Posts: 234 Credit: 15,020 RAC: 0 |
There are two different issues: I agree too. And as for having the uneven playing field here, this is the reason for that I actually have started to consider if I'll be here after my 15,000 milestone here. [b]"I'm trying to maintain a shred of dignity in this world." - Me[/b]
|
|
carl.h Send message Joined: 28 Dec 05 Posts: 555 Credit: 183,449 RAC: 0 |
David Baker has said No backdating, partly I think because of the work involved. I`ll take that as Gospel and not wish for dreams to come to true. Given that what are the options? 1) Zero out and start again. 2) Give the new system a period of time to equalize and go completely over to them. 3) Carry on as we are with the new system being implemented on top of the old at some given point. Any others that have not been ruled out by the project devs already ? Not all Czech`s bounce but I`d like to try with Barbar ;-) Make no mistake This IS the TEDDIES TEAM. |
Astro Send message Joined: 2 Oct 05 Posts: 987 Credit: 500,253 RAC: 0 |
John, when considering whether or not to back date (which they've said they won't), they need to consider who might leave just because they back dated. I'd "assume" the users most likely to do so would be those using optimized core clients. I don't know if you saw it, but I decided to get a random sample of rosetta users and see how many people actually posted or not. I used this base to also find out how many active hosts they had, and which Boinc version they were using. I found that most (a very high percentage) used and continue to use whichever boinc version was standard at the time they signed up. I guess it makes sense, since only 4% have ever posted and just over 1% have ever posted more than once. They just don't see that there's an updated version of boinc available, let alone the existence of third party apps. I selected 5 groups for study. They are Group 1 (userID 10,000-10,250), Group 2 (userID 30,000-30,250), Group 3 (userID 50,000-50,250), Group 4 (userID 70,000-70,250), and Group 5 (user ID 90,000-90,250). Below is a chart showing how many and what percentage apply to their status.  As you can see, out the possible 1,250 userid slots, only 893 are actually filled. 382 are Active(42.78%), 452 are Inactive (50.62%), and 56 are hidden (6.27%). The definition of active is they have a puter attached, and not necessarily that they have any results attributed to them. I.E they aren't removed from Rosettas list until after 30 days of no results. After going through each host on each user, I find that there are only 357 active hosts in groups 1-5. Active hosts means "has a resultID I can check to determine which BV they use. The following is the breakdown of boinc versions used by those hosts.  . |
|
carl.h Send message Joined: 28 Dec 05 Posts: 555 Credit: 183,449 RAC: 0 |
Can`t resist can you Castro...and off we go again. Well done mate, bloomin well done ! And that say`s what about today and tomorrow ? Nothing ! You`re just backdating the row ! Your info means squat and has not been verified neither are most interested. How many XS users post ? Teddies ? Very few they let one or two talk for the great many. Not all Czech`s bounce but I`d like to try with Barbar ;-) Make no mistake This IS the TEDDIES TEAM. |
Vester Send message Joined: 2 Nov 05 Posts: 258 Credit: 3,651,260 RAC: 428 |
Why not start a new stats data base and call the old stats table Legacy Statistics? It has been done elsewhere. ...the Rosetta Legacy Statistics... |
|
Astro Send message Joined: 2 Oct 05 Posts: 987 Credit: 500,253 RAC: 0 |
I will send the complete file to anyone. I have provided links for your use, You'll have a much easier time of it than I had creating it. The complete file includes every user, every host, and every result ID has a link, which you can easily verify the data with. Find a flaw in the methodology used to create it. Just because you don't like the facts, doesn't mean they aren't true non the less, and screaming about them, won't change it either. If you want to dispute these numbers, let me know, I'll be happy to oblige. |
|
Astro Send message Joined: 2 Oct 05 Posts: 987 Credit: 500,253 RAC: 0 |
Please be a bit more specific. What is wrong with the sample? I fully admit it is what I said it is, and that is a sample. It's a evenly distributed random cross section of this population. |
|
carl.h Send message Joined: 28 Dec 05 Posts: 555 Credit: 183,449 RAC: 0 |
It, imo, does not accurately reflect the whole and as a statistician you must realise that, where are your error margins what do you think they are likely to be. To project this as a true representation of Rosetta is foolish and inaccurate. It is no more accurate than the papers who phone 100 people at random to ask which way they are voting. Not all Czech`s bounce but I`d like to try with Barbar ;-) Make no mistake This IS the TEDDIES TEAM. |
|
kevint Send message Joined: 8 Oct 05 Posts: 84 Credit: 2,530,451 RAC: 0 |
John, You have more posts in the last couple of days than credits / WU crunched, why is this an issue with you. SETI.USA  |
|
m00kie Send message Joined: 11 May 06 Posts: 1 Credit: 97,222 RAC: 0 |
Here we go again. Hasn't Dr Baker already weighed in on this? This may not yet be like beating a dead horse but that horse is on its last legs. How many of these threads do we have to have? Proud member of Team Starfire World BOINC |
Message boards :
Number crunching :
One last request for backdating of credits.
©2024 University of Washington
https://www.bakerlab.org