Cheating on Formula BOINC - The Scottish BOINC Team

[Icecold]

For anybody that doesn't closely follow the Formula BOINC forum, some drama is unfolding over at Formula BOINC currently. I'm sure it will come as no surprise to anybody here who has competed against The Scottish BOINC Team (TSBT) before and seen the various ways they act unscrupulously(having been caught blatantly cheating multiple times, "banished" from projects, abused their role in the sprint committee on Formula BOINC, etc.), but they just got caught during the latest Formula BOINC sprint totally inflating their points by changing the benchmark score of their machines so that they would get drastically higher points for each work unit crunched than they should have. Basically, they are unable to beat Planet 3D Now! in a fair competition, and resorted to 100% flat out cheating. They were caught by the admin of SiDock who took all their credit back (reflected on Free-DC) on the project, but it remains to be seen if Formula BOINC will do anything about it to make the sprint right.

What's truly interesting is that it was pretty much every single member of TSBT that was competing in the sprint, so it's not just 1 or 2 bad apples the entire team is incapable of winning without resorting to blatant cheating.

Project Team Stats for The Scottish Boinc Team in SiDock@home

Sidock admin post about it - https://www.sidock.si/sidock/forum_thread.php?id=292#2352

I for one plan on running the next Formula BOINC sprint(Primegrid) on behalf of Planet 3D Now! and will be switching my team affiliation to them on Primegrid, to help ensure that they get first place in Formula BOINC since they were the victim of blatant cheating and poor sportsmanship and deserve to win first place against TSBT. Hopefully Tim over at Formula BOINC adjusts the sprint credits since they clearly should have lost, but based on his response and banning @Orange Kid for posting about it, it's not looking too good. I know this goes without saying, but I'm very proud of the fact that our team always crunches ethically and doesn't try to game the system or do any of that kind of stuff.

 

[Skillz]

Well this doesn't surprise me one bit. Same ole crap they've been doing for years.

Just glad a project admin actually did something about it.

 

[cellarnoise]

How did they do this? Is this a "Credit New" project?

What can be done to fix this on this project and others that use point systems like this?

Crazy how many points looked to be scammed!

Besides helping out the science mission on a project like this, which I have been running among my top 5 for months now, why include them in a "competition"?

I also wonder what else is going on to cheat the projects included in the Boinc competitions?

I find it fun and difficult enough to try and get my hardware to run most efficiently without trying to modify the very point system that the projects are using?

 

[mmonnin03]

Looks like pututu at H as well

 

[cellarnoise]

I've only been around for a few years, but I don't think I have seen this before from Free-DC? Does not look good!

 

[cellarnoise]

I'm going to log out of Free-DC and maybe log back in tomorrow. This kind of feels dirty....

 

[cellarnoise]

Ouch, can I share this here? I'm still learning how to share screen shots or whatever and my freeee-dc link did not work... It does seem that at least the Scottesh Boincer team was doing something to the point system? Maybe I just stop running Sidock for some time until they get this figurued out?

Is this project used for GridCoin results also? How many other things are impacted by inflated points?

 

[Icecold]

@cellarnoise I wouldn't avoid running Sidock, the admins found the problem and fixed it hence the negative credits for all of those members that were cheating the benchmark score. If anything that should give you a lot more confidence that the Sidock admins are dedicated to running a project that awards points fairly. I guess the question would be what other projects exist without that safeguard in place.

 

[In2Photos]

What exactly do you win in these competitions?

 

[Icecold]

What exactly do you win in these competitions?

No actual award other than knowing that you won, and any feeling of accomplishment or prestige that comes with it. Some people take the competitions very seriously and have invested $20,000-$30,0000 or more in hardware specifically for BOINC and Folding@home.

 

[In2Photos]

No actual award other than knowing that you won, and any feeling of accomplishment or prestige that comes with it. Some people take the competitions very seriously and have invested $20,000-$30,0000 or more in hardware specifically for BOINC and Folding@home.

So they're cheating in a competition where the only award is bragging rights? That's crazy!

 

[Icecold]

So they're cheating in a competition where the only award is bragging rights? That's crazy!

Yeah especially when your bragging rights don't mean anything because you're known by everybody in the community as shady people who cheat in the competitions 🤣

 

[StefanR5R]

How did they do this? Is this a "Credit New" project?

I think SiDock is using Credit New but am not sure. They also have a minimum quorum of 1 and initial replication of 1 of the workunits. (I don't know if that's all workunits, or most but not all workunits. The latter method is known as adaptive replication and is also employed by several but not all World Community Grid projects. I may misremember, but I believe SiDock switched to adaptive replication at some point; not sure if they had and if they still have it in place.) In case of varying and unpredictable workunit sizes (which is a reason to rely on an estimation scheme like CreditNew instead of giving predetermined credit), if each workunit or most workunits only receive a single result, then the server cannot directly relate performances of different hosts for its credits estimation. The algorithm is then much more vulnerable to mispredictions than it is with a quorum of 2.

What exactly do you win in these competitions?

There have been Distributed Computing competitions occasionally in which actual prizes could be won (gifted by the initiator of the contest, or donated upfront by some of the community members themselves who were participating), but these are rare exceptions. The Formula BOINC contests in particular don't have any material prizes. The real reasons for such contests are (ideally):
– being a friendly virtual get-together of (parts of) the Distributed Computing community, with the mentioned aspect of bragging rights mixed in;
– to motivate donors of computer time to step it up a notch or two;
– to improve the visibility of Distributed Computing; e.g. one of AnandTech's past calls to Folding@Home races happened to suck myself into this hobby;
– in case of the Formula BOINC series of contests, its main goal back at the time when it was conceived was to draw some attention of the Distributed Computing community towards the less visible and less popular projects who are more starved for donated computer hours than the bigger, more established projects.

 

[Icecold]

Quick update - the latest post on the Formula BOINC forum states that they are going to discuss it and that some type of action will be taken, so hopefully something adequate happens to fix the points earned in the last sprint.

 

[Skillz]

Oh the irony in that statement.

 

[mmonnin03]

So they're cheating in a competition where the only award is bragging rights? That's crazy!

Yeah internet cookies for the most part.

If they are setup to earn GRC they could transfer a tiny, tiny bit of those internet cookies to BTC/$. But cheating to do so is a good way to get the project removed from earning any GRC at all.

 

[Skillz]

So allegedly TSBT has resigned from FB for the 2025 year.



Nothing else has been done to correct the Sprint they cheated in.

I also bet TSBT just signs up for FB using their alt team, aka the Anguillan Pirates team.

So they'll still be participating, but just under a different name.

 

[mmonnin03]

Formula BOINC (Page 3) - EVGA Forums

Looks like there will not be a Formula BOINC for 2024. The guy who runs it does not have the time to do so.Plus there were attacks on the server during sprints. Sad to see it go. Hope he lets Tim take it over but i would not bet on that.Time will tell.

forums.evga.com

Oh it is a 5 min job to fix it. Just change the number in the db to 0.

I'd rather have Ian&Steve as a rep than Timbo

I didn't know there was a 25th Sprint this year, an extra beyond the formal F1 season and I'd guess that came about fairly recent with the update on Willy.
I also don't know of a way an eternal site can gather data to limit stats the way PG can. They could limit it by all GFN tasks at best but that would include all tasks downloaded before hand. Another site cannot scrape data to see when tasks were downloaded, esp not at PG with the main/check link disabled until both completed. Anyone hiding host data would get no credit by eternal gathering of user data.

Seb would basically just be copy/pasting the team state number from PG but then why have a copy of another event.

Challenge statistics

 

[Skillz]

PG exports the stats they collect for their challenge series.

Just go to any one of the individuals or team pages with stats. Replace the html in the URL with .xml.gz and you'll have a nicely formatted xml document of all the stats to do with as you need/want.

 

[StefanR5R]

I was a bit surprised to learn this morning that suddenly a (subset of a) PrimeGrid challenge doubles as a Formula BOINC sprint now. (I haven't been watching Formula BOINC at all for a long time now.)

Seb would basically just be copy/pasting the team state number from PG

When Sébastien implemented this, he hopefully recalled that PrimeGrid challenges have *two* end dates: The predetermined date at which submission of results end (but credit is still only preliminary), and an undetermined date at which validations end (and that's when credit gets canonical). All the time until the latter date, subtractions from preliminary credit are possible. The period between these two dates is also known as the cleanup stage of a PrimeGrid challenge.

Or one could say that there are even four ends (three preliminary, one real), in this order: The predetermined date at which result submissions end, an undetermined date at which validations advanced enough that team positions get canonical, an undetermined date at which individual positions get canonical, and finally an undetermined date at which credits get canonical.

Fortunately, the current PrimeGrid applications, genefer22g included I believe, have quite strong self-checks so that miscalculated results are unlikely to make it into a result report from client to server. Furthermore, the "fast proof" mechanism has generally shortened the cleanup stage a lot, compared to earlier years when a quorum of 2 was needed for validation.

For example, the GFN18/19/20 challenge of December 10th-20th 2023 had its team rankings finalized on December 25th (source) and credit finalized on December 28th, without any subtractions at all (source) edit: had its team rankings finalized on December 23rd and individual rankings finalized on December 28th, without any credit subtractions at all as far as I have seen.

Indeed, credit subtractions during PrimeGrid challenges, cleanup stage inclusive, are extremely unlikely these days, thanks to the mentioned self-checks by the PrimeGrid applications. _And_ because participants in PrimeGrid challenges actually do run PrimeGrid's applications (or if they made their own binaries, their aim is to get results which will pass validation). But this is a Formula BOINC sprint now. And who knows what some more creative folk among the Formula BOINC participants might come up with.

Of course, if Sébastien didn't implement the cleanup stage yet, it's still possible to react in retrospect if necessary. On the other hand, if credit subtractions which happened at SiDock after the fact were not manually imported, then who knows if PrimeGrid credit subtractions would be.

However, if Formula BOINC sprint participants believe that Sébastien implemented or will at least retrospectively act on PrimeGrid's cleanup stage, then these participants will of course have little reason to produce results with anything else than PrimeGrid's very own application binaries.

 

[Icecold]

Indeed, credit subtractions during PrimeGrid challenges, cleanup stage inclusive, are extremely unlikely these days, thanks to the mentioned self-checks by the PrimeGrid applications. _And_ because participants in PrimeGrid challenges actually do run PrimeGrid's applications (or if they made their own binaries, their aim is to get results which will pass validation). But this is a Formula BOINC sprint now. And who knows what some more creative folk among the Formula BOINC participants might come up with.

Of course, if Sébastien didn't implement the cleanup stage yet, it's still possible to react in retrospect if necessary. On the other hand, if credit subtractions which happened at SiDock after the fact were not manually imported, then who knows if PrimeGrid credit subtractions would be.

However, if Formula BOINC sprint participants believe that Sébastien implemented or will at least retrospectively act on PrimeGrid's cleanup stage, then these participants will of course have little reason to produce results with anything else than PrimeGrid's very own application binaries.

I wouldn't put it past somebody to try to game the system and use a custom app that doesn't produce valid results (but produces a lot of points) but man would it be a bad idea for them to try it at Primegrid 🤣 Primegrid runs such a tight ship(thus would quickly identify any hosts with a high validation failure rate) and are so well known and respected that if somebody was cheating there and the PG admins noticed(which they absolutely would) the uproar surrounding it would be huge and the reputation of the cheaters would be ruined even worse than what happened at Sidock. Not that people that are looking to game the system necessarily think that far ahead, but they would really be messing up if they tried that at Primegrid.

 

[StefanR5R]

The musings behind my long-winded posts were motivated by the question "how to copy+paste PG challenge stats technically correctly", not really by the question of "how to participate creatively in an FB sprint which is based on a PG challenge". ;-) . . . The length and depth of my post got out of hand while writing and are not indicative of how much FB concerns me personally these days. :-D

 

[Markfw]

I wouldn't put it past somebody to try to game the system and use a custom app that doesn't produce valid results (but produces a lot of points) but man would it be a bad idea for them to try it at Primegrid 🤣 Primegrid runs such a tight ship(thus would quickly identify any hosts with a high validation failure rate) and are so well known and respected that if somebody was cheating there and the PG admins noticed(which they absolutely would) the uproar surrounding it would be huge and the reputation of the cheaters would be ruined even worse than what happened at Sidock. Not that people that are looking to game the system necessarily think that far ahead, but they would really be messing up if they tried that at Primegrid.

We may need you on the PG challenge. I thought I would be more in the lead, but obviously not. Please come back to TEAM Anandtech as soon as possible.

 

[Icecold]

We may need you on the PG challenge. I thought I would be more in the lead, but obviously not. Please come back to TEAM Anandtech as soon as possible.

I'm not going to switch teams mid challenge and already committed to running this challenge for P3D. But Team Anandtech has an over 20 million point lead from the next closest team so I'm sure you guys will be fine.

 

[TennesseeTony]

I guess I'll be the outlier here, and say Rubbing is racing, All is fair in love and war, etc.

Cloud computing
Running multiple instances to get around 1000 task/day limits
Custom managers to get around 1000 task/day limits
Scripts that update update update, and steal all the GPU Grid and WCG GPU tasks

All circumvent the "fair" policies.

Heck, I'm kinda impressed they found the loophole in the first place. But I do agree with their punishment, all suspect points removed from the project. But perhaps they should be returned at calculated "fair value", as they DID complete those tasks.