Future of the Formula BOINC

[bill1024]

I think they really have to put a message right on the stats page in big red letters. Maybe right in the banner on top.
https://formula-boinc.org/index.py?lang ... &year=2021
I bet a lot of people just check stats and never read the forums so how else will they know.

Then again people are busy this time of year with the holidays, getting ready for Winter and so on.
Suppose if not many sign up go back to the way it was without registration but keep other changes?
Worse comes to worse keep it the way it was.

[pschoefer]

Better late than never, a few thoughts from a member of a team that had lost any interest in the Formula BOINC years ago (I could not even find a replacement for myself on the former Steering Committee when I wanted to resign, but that committee never became active again anyway), but came back to participate a bit more seriously this year.

Originally, we were planning to collect a few points in the sprints here and there to avoid slipping even further in the overall ranking, but when it became clear that we could pull off the 3rd place overall if we performed solidly in the sprints, we ended up putting in a lot more effort. However, with up to 6 sprints in 7 weeks and speculative pre-bunkering in between, this sucked up too much oxygen for a large team with a broad range of different interests. So while we signed up for next year, I won't be surprised if we participate much less actively again.

The sprints really add something to this competition (and effectively address my old concern that all the action was concentrated to the final few weeks), but they need a lot of attention if you want to do well. Sure, you could skip a few sprints, but as soon as you start more or less ignoring some sprints, you'll lose many points (that's how L'AF lost an easy 3rd place this year) and have less incentive to do better in the other sprints. While I understand the announced rule changes for next year from an organizer's point of view (although I think that announcing the project only at the beginning of the actual sprint creates more problems than it solves), they'll require even more effort from the participants to use their hardware most effectively.

A significantly smaller number of sprints would also mitigate the problem that there are either a lot of repeating projects or it is too easy to predict the project and pre-bunker accordingly. Yes, this would break the Formula One analogy, but, honestly, who cares? If the Formula One analogy would be taken seriously, the still ongoing 2021 season would have finished two weeks ago and the 2022 season would not start until March. In addition, it might be a good idea to announce the results of the sprint committee vote only at the end of the sprint season rather than immediately with the project announcement for each sprint, as it was very easy this year to see certain patterns and come up with a good guess for the next project. For example, both L'AF and we had prepared notable bunkers on the projects that received 1-0 votes for the last sprint - except for GPUGRID (almost no tasks available in the last several months, so a very odd choice anyway) and MilkyWay@home (most cumbersome for bunkering, so probably the best possible project from that shortlist).

Speculative pre-bunkering is one of my major annoyances with volunteer-organized competitions such as the Pentathlon and the FB sprints. Better hardware, more aggressive bunkering techniques, and a smaller pool of projects to choose from made this worse in recent years, up to the point that it breaks some things at projects that were once believed to be a picture of stability, and annoys other volunteers who are interested in those projects but not in competitions. It also makes the competitions more boring because it usually just scales up the differences in computing power. And it leads to an unlevel playing field, as there is always someone who knows the project in advance and therefore can't engage in speculation (we have seen in that disastrous ODLK sprint what happens otherwise).

Unfortunately, the problem is hard to solve. You can write what you want in the rules, but unless you can actually enforce the rules, it only takes a few participants who break the rules to turn everything into a shitshow. So the only real solution is on the server side and thus requires significantly more buy-in from project administrators. The same is true for other major annoyances such as tasks not validated during the competition and team hopping (ok, this particular problem could be mitigated by counting the cumulative credit of registered users instead of simply using the team stats, but user-level stats are another can of worms if you think about them carefully) - and I don't even want to talk about several projects granting credit in horribly inconsistent ways that hinder a fair competition. Some project administrators might be willing to help a bit, but most if not all of them don't have the time (and some don't have the coding skills) to develop a scalable, enhanced version of what PrimeGrid does for their challenge series. I could keep writing for quite some time to lay out my ideas on this topic (that somewhat depresses me because I also lack the skills and know-how to solve the problems in practice), but that would be more about the future of Volunteer Computing competitions in general rather than the Formula BOINC, so I'll stop here.


A more concrete piece of feedback: It can be hard to follow where a team recently gained or lost points, and where it is catching up or losing ground, as the website provides only the latest numbers. The teams that have been actively participating in the Formula BOINC for many years probably have their tools in place to keep track of how they are doing, but for a new team, this might be a big obstacle. The "last xx hours" columns that were added to the sprint stats during the final sprint were a very welcome addition. I agree with David that it would be great to have something similar in the marathon stats.

And finally a comment regarding the "small number of registrations": As the number of active volunteers decreased a lot in the last decade, a lot of teams also died. Quite a few teams are only active at one project or a small subset of projects, so there's no real reason for them to sign up for a multi-project competition. Some other teams are not interested in competitions at all. From my experience with the Pentathlon, I'd say that there is a hard core of 15-20 teams that are always in and almost all of them also signed up for next year's Formula BOINC. Then there is a pool of less active teams or sporadically active one-member teams, and a few new teams every now and then; the longer you leave the registration open, the more of those will enter. For the Pentathlon, we usually open the registration 3-4 weeks before the start and end up with around 30 registered teams, so the 23 teams (+2 waiting for confirmation) registered right now look reasonable to me.

[Cougarpelardou]

Nice analysis

[bluestang]

Some good info/advice there @pschoefer that should be taken in by the FB Team.

I think there are far too many Sprints as well and too easy for speculative bunkering for reasons like you mentioned.

Also, I like the expanded stats that were added to the last Sprint...really nice!

[marsinp]

Hello everyone.
2021 was dificult. We crunch as much as possible.
I see the registered team : very few. It show few interest on FB. It is pitty (sad).
But seem a reality.
Thank you to SEB and TIM to keep it.
Now, the goal opf my post :
Happy new year, take care about you and familly..
As joke : Please understand it as a joke) All our power on one project to find the solution against Covid !
We can do it. All together, we have computation power.
The human need to find a solution.

Best regards everyone, Happy new year

[UBT - Timbo]

I think there are far too many Sprints as well and too easy for speculative bunkering for reasons like you mentioned.

Hi bluestang

The "problem" (if there is one) is that the Formula BOINC season, follows the Formula One season...so if a race is held on a weekend, then that is the date used for the Sprint weekend.

And if Formula One is increasing the number of scheduled races (per season) so we will have more Sprints.

To change away from this schedule, would make FB not the same as Formula One...in which case, other sports could be used (as a model) that have fewer "events".

As for speculative bunkering...it wasn't so easy to guess which project was chosen (during 2021) as any one of the projects could be chosen from the full list...so in some cases a project was chosen twice, in order to keep it as fair as possible.

As you may know, in previous years, towards the end of the season, there was only one or two projects that could be chosen, so bunkering was more easy to do.

Maybe the trick for 2022 is NOT to publish a list of Sprint projects - then it will be a surprise for everyone ?

regards
Tim

[pschoefer]

To change away from this schedule, would make FB not the same as Formula One...in which case, other sports could be used (as a model) that have fewer "events".

Please keep in mind that the sprints were only introduced in 2017. In the 10 seasons before, FB and F1 had a different number of races (there were far more FB projects than F1 races back then), the FB races all ran in parallel rather than only one at a time, and there was (and still is) no nice long winter break between FB seasons (only one month break for a couple of years and no break at all for most of the time). The only thing FB and F1 had in common was the points scoring system. Moreover, in 2 of the 5 seasons since the introduction of the sprints (2020 and 2021), the FB sprint schedule did not follow the actual F1 race schedule but only some tentative schedule from months before the start of the season. So, for 80% of its history, FB did not mirror F1 all that closely, and still I have never seen loud complains that the name would be a poor choice.

Yes, it might be nice to follow a real-world model, but sometimes it just works better if you deviate from that model. For example, in a real-world javelin throw competition, you have six tries and only the best throw counts. In the Pentathlon version of it, there are only five tries (simply because we like the number 5 better ) and the third-best throw counts. Of course, we could easily do six tries and count the best throw only, but this would lead to everyone bunkering as much as the deadline permits for one single try and ignore the other tries, so our adoption is obviously more interesting than a plain implementation of the real-world model would be.

On a different note, please also note that F1 introduced "sprint races" last season. In their world, a sprint is a different qualifying format, gives far less points than the real race (3, 2, 1 points for the first three places), and there are only six sprint races planned for the whole 2022 season. So if you really wanted to follow the real-world model letter by letter, you would either have to come up with a new name for the FB sprints or have only six FB sprints that give almost no points (so this would actually be a step back to the pre-2017 FB with only marathons and all the action concentrated in the final few weeks).

As for speculative bunkering...it wasn't so easy to guess which project was chosen (during 2021) as any one of the projects could be chosen from the full list...so in some cases a project was chosen twice, in order to keep it as fair as possible.

As you may know, in previous years, towards the end of the season, there was only one or two projects that could be chosen, so bunkering was more easy to do.

Maybe the trick for 2022 is NOT to publish a list of Sprint projects - then it will be a surprise for everyone ?

If you just exclude projects that obviously can't handle a Sprint (like ODLK, TN-Grid, or WEP-M+2), there's no point in keeping the list secret, as everyone who is paying attention can create the list themselves. If there were any projects that worked smoothly in the past but now are excluded because of a negative or missing response from the project admin, it might make sense not to make this public. I'd definitely keep the sprint committee's voting results secret until after the last sprint project is announced.

[davidBAM1]

Scrap the voting completely IMHO - the Sprint Committee should exclude projects incapable of supporting a sprint, then exclude the last 3 sprint projects, then it is picked purely at random.

Ideally League 1 would sprint on a different project from the other leagues

[mmonnin]

To change away from this schedule, would make FB not the same as Formula One...in which case, other sports could be used (as a model) that have fewer "events".

Please keep in mind that the sprints were only introduced in 2017. In the 10 seasons before, FB and F1 had a different number of races (there were far more FB projects than F1 races back then), the FB races all ran in parallel rather than only one at a time, and there was (and still is) no nice long winter break between FB seasons (only one month break for a couple of years and no break at all for most of the time). The only thing FB and F1 had in common was the points scoring system. Moreover, in 2 of the 5 seasons since the introduction of the sprints (2020 and 2021), the FB sprint schedule did not follow the actual F1 race schedule but only some tentative schedule from months before the start of the season. So, for 80% of its history, FB did not mirror F1 all that closely, and still I have never seen loud complains that the name would be a poor choice.

Yes, it might be nice to follow a real-world model, but sometimes it just works better if you deviate from that model. For example, in a real-world javelin throw competition, you have six tries and only the best throw counts. In the Pentathlon version of it, there are only five tries (simply because we like the number 5 better ) and the third-best throw counts. Of course, we could easily do six tries and count the best throw only, but this would lead to everyone bunkering as much as the deadline permits for one single try and ignore the other tries, so our adoption is obviously more interesting than a plain implementation of the real-world model would be.

On a different note, please also note that F1 introduced "sprint races" last season. In their world, a sprint is a different qualifying format, gives far less points than the real race (3, 2, 1 points for the first three places), and there are only six sprint races planned for the whole 2022 season. So if you really wanted to follow the real-world model letter by letter, you would either have to come up with a new name for the FB sprints or have only six FB sprints that give almost no points (so this would actually be a step back to the pre-2017 FB with only marathons and all the action concentrated in the final few weeks).

As for speculative bunkering...it wasn't so easy to guess which project was chosen (during 2021) as any one of the projects could be chosen from the full list...so in some cases a project was chosen twice, in order to keep it as fair as possible.

As you may know, in previous years, towards the end of the season, there was only one or two projects that could be chosen, so bunkering was more easy to do.

Maybe the trick for 2022 is NOT to publish a list of Sprint projects - then it will be a surprise for everyone ?

If you just exclude projects that obviously can't handle a Sprint (like ODLK, TN-Grid, or WEP-M+2), there's no point in keeping the list secret, as everyone who is paying attention can create the list themselves. If there were any projects that worked smoothly in the past but now are excluded because of a negative or missing response from the project admin, it might make sense not to make this public. I'd definitely keep the sprint committee's voting results secret until after the last sprint project is announced.

The FB Sprints and F1 races did not mirror in 2020/2021 because F1 changed their race schedule after it was announced due to COVID restrictions. If the race schedule would not have changed then FB Sprints would have matched the races.

FB Sprint has nothing to do with the F1 Sprints introduced last year. Stop trying to make a comparison.

[pschoefer]

The FB Sprints and F1 races did not mirror in 2020/2021 because F1 changed their race schedule after it was announced due to COVID restrictions. If the race schedule would not have changed then FB Sprints would have matched the races.

I'm aware of this. Still, it is an(other) example for FB not mirroring F1 and nobody being less likely to participate in the Sprints because of this - at least I have not seen any complaints about this.

FB Sprint has nothing to do with the F1 Sprints introduced last year. Stop trying to make a comparison.

Would you write the same if the F1 Sprints had already been in place when the FB Sprints were introduced five years ago? I just want to make the point that it already is completely arbitrary where FB does follow F1 and where it doesn't, so I believe that it doesn't matter if there is a FB Sprint on every F1 weekend, or, for example, only every other F1 weekend, or two FB Sprints in parallel whenever there is a F1 Sprint, or no relation between FB sprints and F1 races at all.

On the other hand, I do believe that having fewer FB Sprints would make the FB more attractive because this would leave more room for other activities without losing too many FB points. Of course, this is only based on my own observations, maybe everyone else is fine with 23 Sprints and it wouldn't attract any additional participants but cause other participants to stop participating if there were fewer Sprints.