Sprints 2024 - timetable and projects selected
Re: Sprints 2024 - timetable and projects selected
One of the SiDock admins made a post about it "SiDock@home is not F1..."
- Attachments
-
- 2024-12-08 16_27_35d-Greenshot.png (536.18 KiB) Viewed 527 times
-
- Posts: 102
- Joined: Fri Dec 25, 2020 6:18 am
Re: Sprints 2024 - timetable and projects selected
Based on this, and I think the data speaks for itself, I guess the question comes up on how stuff like this is handled moving forward. I've been a proponent of our team rejoining Formula BOINC in the past, but I think that there needs to be some ground rules or at least a discussion as far as what is expected from competitors in terms of ethics and rules and what happens if an issue like this is discovered. Formula BOINC is of course not in a position to necessarily police how people are crunching projects(that is up to the projects themselves really), but if I would ever have any chance of convincing the team I'm on to rejoin Formula BOINC it would be helpful to have some assurances that it will be made right if a team(ANY team) is blatantly gaming the system. It's pretty clear what the results of this last sprint were if you look at the stats on Free-DC, and it certainly is not the outcome that is shown in Formula BOINC. Note - I'm not calling out any teams or specific people or making any accusations, this is a general comment and it ultimately affects multiple teams, so hopefully you don't feel the need to ban hammer me out of here Timbodj.walnut wrote: ↑Mon Dec 09, 2024 1:10 am One of the SiDock admins made a post about it "SiDock@home is not F1..."
- UBT - Timbo
- Site Admin
- Posts: 1015
- Joined: Wed Nov 23, 2016 1:21 am
- Contact:
Re: Sprints 2024 - timetable and projects selected
Hi dj and Icecold
Thanks for your postings.
I think the question is "what have TSBT (and HardOCP apparently) done" to upset the SIDock admins?
And how can the SiDock project admin substantiate what might have happened and why have they deducted BOINC credits from these teams?
That is beyond the realm of FB, as all I can do is to suggest a specific project to crunch for (during a Sprint) and hope that this is done fairly and with good intent.
The problem is that I can never know HOW anyone would be able to cheat any BOINC project, into falsifying the BOINC credits they have earned. And if there is a way for this to happen then someone who is in charge of running the project must lock down the project to ensure only accredited uploads are accepted and correctly validated. And that is nothing to do with FB.
As for making things right?
If a project admin could spare the time, to give their reasons as to why specific credits have been deleted, then there is the possibility that Sebastien can edit the Sprint results accordingly. But I would need to check with him, if this is possible.
However: This then opens up the possibility that anyone could dispute how any member (or team) earned their credits during a Sprint or at any other time - and I'm sure that project admins have got better things to do than to be checking thousands of validated work units to see what was going on - but then if their back-office systems were up to scratch and could detect "falsely credited uploads" then this problem would go away?
And no need for ban hammers when members offer constructive, level headed topics for discussion
regards and thanks
Tim
Thanks for your postings.
I think the question is "what have TSBT (and HardOCP apparently) done" to upset the SIDock admins?
And how can the SiDock project admin substantiate what might have happened and why have they deducted BOINC credits from these teams?
That is beyond the realm of FB, as all I can do is to suggest a specific project to crunch for (during a Sprint) and hope that this is done fairly and with good intent.
The problem is that I can never know HOW anyone would be able to cheat any BOINC project, into falsifying the BOINC credits they have earned. And if there is a way for this to happen then someone who is in charge of running the project must lock down the project to ensure only accredited uploads are accepted and correctly validated. And that is nothing to do with FB.
As for making things right?
If a project admin could spare the time, to give their reasons as to why specific credits have been deleted, then there is the possibility that Sebastien can edit the Sprint results accordingly. But I would need to check with him, if this is possible.
However: This then opens up the possibility that anyone could dispute how any member (or team) earned their credits during a Sprint or at any other time - and I'm sure that project admins have got better things to do than to be checking thousands of validated work units to see what was going on - but then if their back-office systems were up to scratch and could detect "falsely credited uploads" then this problem would go away?
And no need for ban hammers when members offer constructive, level headed topics for discussion
regards and thanks
Tim
-
- Posts: 102
- Joined: Fri Dec 25, 2020 6:18 am
Re: Sprints 2024 - timetable and projects selected
Sidock uses credit new. This can be explained better here than I can - https://boinc.berkeley.edu/trac/wiki/CreditNew but the gist of it is that the points for each completed work unit are calculated factoring in the total computing power (flops) of the CPU that did the calculation. BOINC does a CPU benchmark on initial run, and can also be ran through BOINC Manager --> Tools --> Run CPU Benchmarks. This then shows as your 'Measured floating point speed' and 'Measured Integer Speed' on your host on that project site. During the sprint there were hundreds(possibly thousands) of hosts that had been created with a way inflated falsified benchmark score. The file containing the benchmark score had been manually edited to reflect a far higher value than it would contain normally. This resulted in those hosts getting upwards of 7,000-10,000 credits per task, while every other person crunching normally was getting 150 to 200 credits per task.UBT - Timbo wrote: ↑Mon Dec 09, 2024 1:02 pm The problem is that I can never know HOW anyone would be able to cheat any BOINC project, into falsifying the BOINC credits they have earned. And if there is a way for this to happen then someone who is in charge of running the project must lock down the project to ensure only accredited uploads are accepted and correctly validated. And that is nothing to do with FB.
As for making things right?
If a project admin could spare the time, to give their reasons as to why specific credits have been deleted, then there is the possibility that Sebastien can edit the Sprint results accordingly. But I would need to check with him, if this is possible.
However: This then opens up the possibility that anyone could dispute how any member (or team) earned their credits during a Sprint or at any other time - and I'm sure that project admins have got better things to do than to be checking thousands of validated work units to see what was going on - but then if their back-office systems were up to scratch and could detect "falsely credited uploads" then this problem would go away?
And no need for ban hammers when members offer constructive, level headed topics for discussion
regards and thanks
Tim
I personally have looked at 50+ hosts that were getting incorrect credit - most had their NCPU's changed as well as their GPU coproc file spoofed as well, it was clearly done with the intent to game the system. I'm not a project admin though, and the hosts were all for users with their computers hidden, so I have no way of knowing who those specific hosts/teams belonged to.
The Sidock admins then noticed that, and corrected the points for those tasks/hosts to what they should be and have put in guards to prevent it moving forward from my understanding. I think everybody would agree that getting 7,000 points for the exact same work unit somebody else would get 200 points for is not fair and very against the spirit of any competition. The points for the sprint should be manually adjusted for all teams to what the actual credits ended up being, had the admins at Sidock adjusted the credits before the sprint was over that would have happened automatically. Otherwise it gives every team a green light to do anything they can to game the system moving forward.
I don't think doing the right thing to correct points in this specific very clear instance of invalid credits opens up a huge can of worms to people disputing member or team credits after the fact in future competitions. It would also go a long way for people if it would be stated very clearly by Formula BOINC (either you or Sebastien) that cheating is not allowed in Formula BOINC and that crunchers are expected to run Formula BOINC in a fair and ethical manner. The BOINC Pentathlon has this verbiage on their rules page - "Any behaviour that intentionally harms the projects or other teams devalues the Pentathlon and may lead to consequences from the organizers, the other participants or the project administrators." but there doesn't seem to be a similar concise message anywhere in Formula BOINC.
Hopefully that still all falls in the realm of level headed and constructive(thus avoiding the ban hammer), I made sure not to throw out accusations at specific members or teams just state the facts of what happened.
- UBT - Timbo
- Site Admin
- Posts: 1015
- Joined: Wed Nov 23, 2016 1:21 am
- Contact:
Re: Sprints 2024 - timetable and projects selected
Thanks for that - very useful info which is much appreciated.Icecold-TAAT wrote: ↑Mon Dec 09, 2024 2:00 pm Sidock uses credit new. This can be explained better here than I can - https://boinc.berkeley.edu/trac/wiki/CreditNew but the gist of it is that the points for each completed work unit are calculated factoring in the total computing power (flops) of the CPU that did the calculation. BOINC does a CPU benchmark on initial run, and can also be ran through BOINC Manager --> Tools --> Run CPU Benchmarks. This then shows as your 'Measured floating point speed' and 'Measured Integer Speed' on your host on that project site. During the sprint there were hundreds(possibly thousands) of hosts that had been created with a way inflated falsified benchmark score. The file containing the benchmark score had been manually edited to reflect a far higher value than it would contain normally. This resulted in those hosts getting upwards of 7,000-10,000 credits per task, while every other person crunching normally was getting 150 to 200 credits per task. I personally have looked at 50+ hosts that were getting incorrect credit - most had their NCPU's changed as well as their GPU coproc file spoofed as well, it was clearly done with the intent to game the system. I'm not a project admin though, and the hosts were all for users with their computers hidden, so I have no way of knowing who those specific hosts/teams belonged to. The Sidock admins then noticed that, and corrected the points for those tasks/hosts to what they should be and have put in guards to prevent it moving forward from my understanding. I think everybody would agree that getting 7,000 points for the exact same work unit somebody else would get 200 points for is not fair and very against the spirit of any competition. The points for the sprint should be manually adjusted for all teams to what the actual credits ended up being, had the admins at Sidock adjusted the credits before the sprint was over that would have happened automatically. Otherwise it gives every team a green light to do anything they can to game the system moving forward.
So, it seems to me that SiDock should implement a better validation system as they will know from their upload history (of millions of uploaded results) which CPU's are being used and how many credits are being issued to those CPUs. If someone then "games the system" by inflating their benchmark score for a specific CPU, and starts earning say 50 times the credit per task, then this should be flagged immediately by the validation system and hence the result are marked as invalid (and hence no credits are awarded)?
Clearly, the SiDock admins did notice something strange but maybe after the event rather than (perhaps) during it (and that's not a failure of anyone as not everyone works 24/7 and weekends are usually not work days !)
You make a good point and I am all for trying to keep the playing field level...but as FB has nothing to do with any project, so there is no easy way to put in place any process of checking that every member (and every team) credits were earned legally. We can only rely on the project for updating the member and team total credits each day and any subsequent action by a project (to remove credits) could happen at any point in time and by then it might perhaps be too late to do anything with any FB points awarded.Icecold-TAAT wrote: ↑Mon Dec 09, 2024 2:00 pm I don't think doing the right thing to correct points in this specific very clear instance of invalid credits opens up a huge can of worms to people disputing member or team credits after the fact in future competitions. It would also go a long way for people if it would be stated very clearly by Formula BOINC (either you or Sebastien) that cheating is not allowed in Formula BOINC and that crunchers are expected to run Formula BOINC in a fair and ethical manner. The BOINC Pentathlon has this verbiage on their rules page - "Any behaviour that intentionally harms the projects or other teams devalues the Pentathlon and may lead to consequences from the organizers, the other participants or the project administrators." but there doesn't seem to be a similar concise message anywhere in Formula BOINC.
Hopefully that still all falls in the realm of level headed and constructive(thus avoiding the ban hammer), I made sure not to throw out accusations at specific members or teams just state the facts of what happened.
In the case of this Sprint, I do not think that the actual FB points awarded to each and every participating FB member and team has had a serious impact on the respective league tables. But, it does not bode well, if other Sprints could be (or have been) affected and the relevant project will not (or has not) noticed this.
I would hope that any members of FB teams who read this, can confirm that their team has NOT been gaming the BOINC credit earning system and if they have been found to be cheating, that they admit to this and that the team and/or its member(s) request to be removed from any further participation. This can be done via PM to me either here or via any BOINC project.
But it goes without saying that teams that allow their members to cheat will not be allowed to continue on any FB related challenge - so team founders are requested to make sure their team members are all playing fair.
regards
Tim
Re: Sprints 2024 - timetable and projects selected
Is nothing going to be done? Will there be no disqualifications regarding the sprint results? The teams that had their credits removed should be disqualified.
The SiDock post I linked to even had the admin specifically referencing this contest: 'SiDock@home is not F1.'
This whole situation makes Formula BOINC look terrible.
The SiDock post I linked to even had the admin specifically referencing this contest: 'SiDock@home is not F1.'
This whole situation makes Formula BOINC look terrible.
- UBT - Timbo
- Site Admin
- Posts: 1015
- Joined: Wed Nov 23, 2016 1:21 am
- Contact:
Re: Sprints 2024 - timetable and projects selected
Hidj.walnut wrote: ↑Mon Dec 09, 2024 5:33 pm Is nothing going to be done? Will there be no disqualifications regarding the sprint results? The teams that had their credits removed should be disqualified.
The SiDock post I linked to even had the admin specifically referencing this contest: 'SiDock@home is not F1.'
This whole situation makes Formula BOINC look terrible.
This issue has only surfaced in the last 24 hours and as such, I need to chat with Sebastien as to what action should be taken. And please bear in mind that I am not a 24/7 employee of FB and I have a real job and I have many responsibilities to manage every day.
But it is clear from what I have seen that certain BOINC accounts have falsified their credits which has prompted the SIDock admin to nullify some of the credits earned over the last few days.
This WILL lead to further action being taken and in due course this will be notified on this forum.
regards
Tim
Re: Sprints 2024 - timetable and projects selected
In my understanding it was done this way:
"The way this was happening was the hosts were faking their Flops rating for the CPU (this is easily done by editing the flops rating and then disabling CPU benchmarks in the BOINC client so that it stays inflated and not overwritten by a benchmark run)."
Now the big question is are there other projects that calculate points in that manner, were they used as sprints this year and do the points look suspicious?
If you have to cheat to make yourself feel/look good it really makes you look really really bad. And I am being nice there.
"The way this was happening was the hosts were faking their Flops rating for the CPU (this is easily done by editing the flops rating and then disabling CPU benchmarks in the BOINC client so that it stays inflated and not overwritten by a benchmark run)."
Now the big question is are there other projects that calculate points in that manner, were they used as sprints this year and do the points look suspicious?
If you have to cheat to make yourself feel/look good it really makes you look really really bad. And I am being nice there.
- UBT - Timbo
- Site Admin
- Posts: 1015
- Joined: Wed Nov 23, 2016 1:21 am
- Contact:
Re: Sprints 2024 - timetable and projects selected
Hibill1024 wrote: ↑Mon Dec 09, 2024 8:08 pm In my understanding it was done this way:
"The way this was happening was the hosts were faking their Flops rating for the CPU (this is easily done by editing the flops rating and then disabling CPU benchmarks in the BOINC client so that it stays inflated and not overwritten by a benchmark run)."
Now the big question is are there other projects that calculate points in that manner, were they used as sprints this year and do the points look suspicious?
If you have to cheat to make yourself feel/look good it really makes you look really really bad. And I am being nice there.
Many thanks for your post.
Yes, I understand both of your points and clearly, in the case of the SIDock project, cheating has taken place and BOINC credit adjustments have been made by their admin so, this is really an open and shut case and anyone caught by this should be aware of the consequences of their actions.
Sebastien, the Formula BOINC founder, has been briefed on these events and I will enter into discussions with him as to what action should be taken as such cheating clearly belittles and undermines the honest work of the majority, who want to help various projects by their donation of CPU/GPU cycles. And just plain cheating the credits earned serves NO real or justifiable purpose. In fact it is just childish of such people to "game the system" for the sake of a placement on the FB League tables.
Such action will be made known to all forum members in due course.
EDIT: Of course it is probable that more than one team and more than one member have hacked their benchmarks (in the past) so that they can be awarded more credits per task, compared to what honest crunchers would earn. But it is patently obvious that doing so for the sake on a position on a FB League table is quite contrary to how the majority of people would act. And there's no real benefit in doing this except on one page on one website ! And even if it raises your position by 10, 100, or even 1,000 places, this is nothing to brag about if it was achieved by cheating.
In athletics, if you cheat you get banned from the sport and your results are removed from the history books and any medals you've won are rescinded and other clean athletes get their placings upgraded. Other sports are doing the same, such as cycling.
So, if you can live with yourself by cheating for a few extra BOINC credits, then all I can say is that you must live a lonely and solitary existence if this is how you want to act. It's actually quite pathetic
regards
Tim
Re: Sprints 2024 - timetable and projects selected
Everybody is now (logically) thinking : if *this* was done by *that* team on this particular project, and we can see for a long time that this team has been grossly dominating the competition (at least the sprints, but not only), it is no surprise that we all think that this team probably had a similar behavior on other projects.
But of course it requires technical skills and investigation time to find elements that could lead to further conclusions, that I will not make.
But of course it requires technical skills and investigation time to find elements that could lead to further conclusions, that I will not make.