Another solution for the credit issue that hasn't been mentioned.

Message boards : Number crunching : Another solution for the credit issue that hasn't been mentioned.

To post messages, you must log in.

Previous · 1 . . . 4 · 5 · 6 · 7

AuthorMessage
Profile carl.h
Avatar

Send message
Joined: 28 Dec 05
Posts: 555
Credit: 183,449
RAC: 0
Message 23922 - Posted: 20 Aug 2006, 22:16:17 UTC
Last modified: 20 Aug 2006, 22:19:09 UTC

Yep there is a quorum but over 4 days of crunching it should start to level, do you not agree ?

I have members of my team who`ve been there longer who`ve experienced are experiencing the same. Some with 5.5 some not, same result.

If you wish I`ll try a little experiment with the same machine over a given period with every Boinc project, would you expect parity ?
Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.
ID: 23922 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile carl.h
Avatar

Send message
Joined: 28 Dec 05
Posts: 555
Credit: 183,449
RAC: 0
Message 23931 - Posted: 20 Aug 2006, 22:52:07 UTC

I would be pretty certain there is no parity across Boinc therefore the x project defence is a misnomer !
Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.
ID: 23931 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1831
Credit: 119,627,225
RAC: 9,274
Message 23958 - Posted: 20 Aug 2006, 23:28:07 UTC
Last modified: 20 Aug 2006, 23:32:54 UTC

Gents (and VV ;D)

I've been away for the weekend, and have just read the thread in its entirety (one of the more rational threads over the last week or two!). I believe I've taken a pretty unbiased view of the posts in the thread, and although they aren’t all directly fitting with title of the thread, they are all necessary to put the thread title in context... slippery slope I know Ethan! So I’ve tried to summarise the suggestions:

First, my position on an unrelated suggestion:
Awarding credit on a basis other than purely work done
For me this is very simple - you get credit for, and directly proportional to the work you do. More work = credit. This keeps those into stats happy and healthy competition abundant. I can't see any justification for basing credit on anything other than work done, and any bonus for dedicating more/better resources to Rosetta is given by the increase in work done, and therefore credit received.

Credit options
I think Tallbill's suggestion (although not a new one to me or a few others here!) is a very good one, and I'd be all for it.

Re-assessment of all jobs back to Feb to make them relevant to the work done has been discussed fairly extensively (although not always rationally) so I won’t go into it too much. We know the effects of optimised clients – sometimes their use was intended to level the playing field with the standard client (e.g. on linux machines), and sometimes it was to attempt to level the playing field with regards to others running optimised clients. It is also known that credit assigned to date hasn’t related particularly well to work done – the benchmark isn’t well aligned to work done whether comparing within a group using the standard client or a group using optimised ones. Whetstone + Dhrystone / 2 doesn’t really relate to work done, no matter how many W+D’s your client reports.

If we’d had a fair credit system from the start then there’d have been none of the recent discussions/arguments, but we didn’t. We do now, though (or will very soon hopefully). To move forward, putting the options in context, I believe they are as follows:

1. Zero the stats and start afresh
This isn’t going to be popular by many and I believe many would feel that their contribution to date had been belittled.

2. Start using the new credit system leaving the previous stats as-is
There are a number of people that are for this, and I believe their reasoning is that the rules were as they were. They weren’t breaking any rules and so have legally gained their credits. There is still a choice to make here though - the new credit allocations can be at the rate of the previous standard client, or one of the more popular optimised ones.

3. Re-assessing all jobs processed using the new credit system
The new credit system can be applied to all results processed since early in the project (Feb 06) as it only requires knowledge of the number of decoys produced. Credit allocation per decoy can be calculated automatically from existing data relatively easily.

The result would probably be that those running Linux would see their credits increase as they have been hard-done-by with the current system. I believe P4s would also probably see an increase in credit. Those using optimised clients would probably see a fall in credits. The actual effect on credits isn’t known at the moment, but the result would be that credit would be aligned with work done.

4. The subject of this thread – increasing the new credit system to match optimised clients
As above, but rather than those using optimised clients probably seeing a fall in credit assigned, credit would be assigned to relate to work done, but by increasing the credit proportionately for all users to bring credit allocation up to the level of that claimed by optimised clients. Ranking would be the same as in number 3, but all credit allocations would be higher.

I think I’ve been fair there(?).

And that is what it seems that have been happening here: a ganging up against the power crunchers.

Jose, you posted the above (I hope it doesn’t appear as if I’ve taken the quote out of context), and I’ve picked it out because it’s important to the project. (I'm really not trying to provoke any reaction other than a response, or bring up a matter that's been settled - I don't think it has!)
I'm fairly sure that we all want a system where the power crunchers are shown as such. Any of the proposed credit systems will reward crunchers for the work they've done, so power crunchers will obviously be granted the most credit in any of the following systems. Backdating the credits wouldn't be against the power crunchers - it would/will sit the likes of DDTUNG, PY222 and KevinT squarely at the top of the stats tables where they belong. I don’t see that as ganging up on anyone!

Cross boinc-project equality
Some believe it important, some not, some impossible. I believe it has at least some importance and am fairly certain it's possible. BOINC is a platform that uses credits that are intended to be interchangeable. If cross-project crediting isn’t balanced then there is a risk that a race between projects develops. Projects (especially new ones) will benefit from assigning more credits than other projects in order to attract crunchers. Although as far as I know Game Theory doesn’t account for human decency and altruism very well (maybe J Maynard Smith did?), it tells us that this system isn’t stable. As there is no mediating factor, only the sensibility of the project staff will stop this from escalating, but I'm sure their is temptation there for some, if not all of them.

I therefore believe cross-project credit balancing is important. How easy it is to maintain, and whether it’s more important than, say #3, I don’t know.

Apologies for the long post, but it’s a simple topic with lots of possible viewpoints.

Danny
ID: 23958 · Rating: 3 · rate: Rate + / Rate - Report as offensive    Reply Quote
Ethan
Volunteer moderator

Send message
Joined: 22 Aug 05
Posts: 286
Credit: 9,304,700
RAC: 0
Message 23962 - Posted: 20 Aug 2006, 23:32:12 UTC

. . slipping down slope.

I would say at this point there are no credit options as they relate to past credits. The new system will start working over the top of the current system when it's switched (As per DB).

There is a cross project credit thread, let's keep those comments over there. .

And I'm sure there's something left if your post, let me know if I didn't address it.
ID: 23962 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jose

Send message
Joined: 28 Mar 06
Posts: 820
Credit: 48,297
RAC: 0
Message 23966 - Posted: 20 Aug 2006, 23:37:06 UTC - in response to Message 23958.  

Gents (and VV ;D)

I've been away for the weekend, and have just read the thread in its entirety (one of the more rational threads over the last week or two!). I believe I've taken a pretty unbiased view of the posts in the thread, and although they aren’t all directly fitting with title of the thread, they are all necessary to put the thread title in context... slippery slope I know Ethan! So I’ve tried to summarise the suggestions:

First, my position on an unrelated suggestion:
Awarding credit on a basis other than purely work done
For me this is very simple - you get credit for, and directly proportional to the work you do. More work = credit. This keeps those into stats happy and healthy competition abundant. I can't see any justification for basing credit on anything other than work done, and any bonus for dedicating more/better resources to Rosetta is given by the increase in work done, and therefore credit received.

Credit options
I think Tallbill's suggestion (although not a new one to me or a few others here!) is a very good one, and I'd be all for it.

Re-assessment of all jobs back to Feb to make them relevant to the work done has been discussed fairly extensively (although not always rationally) so I won’t go into it too much. We know the effects of optimised clients – sometimes their use was intended to level the playing field with the standard client (e.g. on linux machines), and sometimes it was to attempt to level the playing field with regards to others running optimised clients. It is also known that credit assigned to date hasn’t related particularly well to work done – the benchmark isn’t well aligned to work done whether comparing within a group using the standard client or a group using optimised ones. Whetstone + Dhrystone / 2 doesn’t really relate to work done, no matter how many W+D’s your client reports.

If we’d had a fair credit system from the start then there’d have been none of the recent discussions/arguments, but we didn’t. We do now, though (or will very soon hopefully). To move forward, putting the options in context, I believe they are as follows:

1. Zero the stats and start afresh
This isn’t going to be popular by many and I believe many would feel that their contribution to date had been belittled.

2. Start using the new credit system leaving the previous stats as-is
There are a number of people that are for this, and I believe their reasoning is that the rules were as they were. They weren’t breaking any rules and so have legally gained their credits. There is still a choice to make here though - the new credit allocations can be at the rate of the previous standard client, or one of the more popular optimised ones.

3. Re-assessing all jobs processed using the new credit system
The new credit system can be applied to all results processed since early in the project (Feb 06) as it only requires knowledge of the number of decoys produced. Credit allocation per decoy can be calculated automatically from existing data relatively easily.

The result would probably be that those running Linux would see their credits increase as they have been hard-done-by with the current system. I believe P4s would also probably see an increase in credit. Those using optimised clients would probably see a fall in credits. The actual effect on credits isn’t known at the moment, but the result would be that credit would be aligned with work done.

4. The subject of this thread – increasing the new credit system to match optimised clients
As above, but rather than those using optimised clients probably seeing a fall in credit assigned, credit would be assigned to relate to work done, but by increasing the credit proportionately for all users to bring credit allocation up to the level of that claimed by optimised clients. Ranking would be the same as in number 3, but all credit allocations would be higher.

I think I’ve been fair there(?).

And that is what it seems that have been happening here: a ganging up against the power crunchers.

Jose, you posted the above (I hope it doesn’t appear as if I’ve taken the quote out of context), and I’ve picked it out because it’s important to the project. (I'm really not trying to provoke any reaction other than a response, or bring up a matter that's been settled - I don't think it has!)
I'm fairly sure that we all want a system where the power crunchers are shown as such. Any of the proposed credit systems will reward crunchers for the work they've done, so power crunchers will obviously be granted the most credit in any of the following systems. Backdating the credits wouldn't be against the power crunchers - it would/will sit the likes of DDTUNG, PY222 and KevinT squarely at the top of the stats tables where they belong. I don’t see that as ganging up on anyone!

Cross boinc-project equality
Some believe it important, some not, some impossible. I believe it has at least some importance and am fairly certain it's possible. BOINC is a platform that uses credits that are intended to be interchangeable. If cross-project crediting isn’t balanced then there is a risk that a race between projects develops. Projects (especially new ones) will benefit from assigning more credits than other projects in order to attract crunchers. Although as far as I know Game Theory doesn’t account for human decency and altruism very well (maybe J Maynard Smith did?), it tells us that this system isn’t stable. As there is no mediating factor, only sensibility will stop this from escalating.

I therefore believe cross-project credit balancing is important. How easy it is to maintain, and whether it’s more important than, say #3, I don’t know.

Apologies for the long post, but it’s a simple topic with lots of possible viewpoints.

Danny


I will not back ANYTHING that involves BACKDATING upwards or Downward. That issue is closed for me. Any Sugestion that involves backdating in any form or degree I will opose. And there is Backdating elements in what you propose thus, I cannot accept what you propose.

As to BOINC: I think my position is now A Fair Rosetta is a BOINC free Rosetta.
ID: 23966 · Rating: -6 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jose

Send message
Joined: 28 Mar 06
Posts: 820
Credit: 48,297
RAC: 0
Message 23967 - Posted: 20 Aug 2006, 23:45:34 UTC
Last modified: 20 Aug 2006, 23:47:04 UTC

Ethan:
The Backdating issue is kept alive by many postings that you allow to stand unchallenged , un-edited . You know the reaction to any backdating rehashing. Why do you keep them visible?

The moment David Baker said NO BACKDATING . That issue should have disappeared from the threads...and yes you keep allowing it . It is as if you are sabotaging David Bakers effort to terminate this backdating stuff.

There is only one way any backdating post in any form is to make them invisible. Do so.
ID: 23967 · Rating: -7 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1831
Credit: 119,627,225
RAC: 9,274
Message 23968 - Posted: 20 Aug 2006, 23:45:52 UTC - in response to Message 23966.  


I cannot accept what you propose.


I don't think i've proposed anything there! I've just summarised to make it clear where Tallbill's suggestion sits in relation what's been discussed to date.
ID: 23968 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jose

Send message
Joined: 28 Mar 06
Posts: 820
Credit: 48,297
RAC: 0
Message 23970 - Posted: 20 Aug 2006, 23:51:41 UTC - in response to Message 23968.  
Last modified: 20 Aug 2006, 23:52:34 UTC


I cannot accept what you propose.


I don't think i've proposed anything there! I've just summarised to make it clear where Tallbill's suggestion sits in relation what's been discussed to date.

'

Tallbill's comment as he developed removed the issue of backdating. You have re-stated the backdating issue. So you changed the prroposal and thus it is your proposal.

So anything that includes backdating is out of the loop. Any attempt to run with it , pass it over or under the radar...will not be considered a serious attemopt at anything.

What is so complicated in David Bakers staement that there will be no backdating that is so hard to understand?

Drat!!!! Stop the rehashing of what has been rejected.
ID: 23970 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1831
Credit: 119,627,225
RAC: 9,274
Message 23971 - Posted: 20 Aug 2006, 23:52:07 UTC - in response to Message 23967.  
Last modified: 20 Aug 2006, 23:59:33 UTC

Ethan:
The Backdating issue is kept alive by many psotings that you allow to stand unchallenged , un-edited . You know the reaction to any backdating rehashing. Why do you keep them visible?

The moment David Baker said NO BACKDATING . That issue should have disappeared from the threads...and yess you keep allowing it . It is as if you are sabotaging David Bakers effort to terminate this backdating stuff.

There is only one way any backdating post in any form is to make them invisible. Do so.


Jose, you seem to be on a one-man quest here (I don't mean you're alone in your opinions, but you're adamant that they're to be taken by all). I don't accept some of your opinions, but I do respect that they are your opinions. I also accept that DB/DK etc will do what they think right for the project and that seems to be no backdating of credit. Fine! My post is still entirely valid and I think it's valid within this thread. I've put what's been discussed into context, which seems fair enough seeing as there are quite a few posts in this thread that are based on misunderstandings.

[edit!]
I've mentioned backdating because it seems to have been THE major issue over the last few weeks (I don't think anyone would disagree with that!) and formed a (the) major part of the thread origin. Whether in conjunction with Bill's original or updated suggestion, I don't think it would be right because, as I stated, I believe cross-project credit balancing is necessary. I don't believe these issues can be considered unless in context

Danny
ID: 23971 · Rating: 3 · rate: Rate + / Rate - Report as offensive    Reply Quote
Ethan
Volunteer moderator

Send message
Joined: 22 Aug 05
Posts: 286
Credit: 9,304,700
RAC: 0
Message 23973 - Posted: 20 Aug 2006, 23:58:40 UTC

Okie Dokie. . the data is valid but since it's been answered (dealing with old credits), this thread is closed but will remain visible (there's nothing wrong with the content).

Please post in other threads regarding other topics as this thread started as a discussion of back crediting.
ID: 23973 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
riptide
Avatar

Send message
Joined: 24 Mar 06
Posts: 27
Credit: 103,422
RAC: 0
Message 24077 - Posted: 21 Aug 2006, 13:22:31 UTC - in response to Message 23832.  

Why am I not allowed to ask Jose why he opposed to a fair view at the past?


You are, you didn't word it that way.

Ethan... i must say your share price is going up in my book!
I love Mr. Smith. He keeps us safe from Alien Scum. He's probably good a Rosetta too.
ID: 24077 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 . . . 4 · 5 · 6 · 7

Message boards : Number crunching : Another solution for the credit issue that hasn't been mentioned.



©2024 University of Washington
https://www.bakerlab.org