Please note that my comments in this post in no way reflect my minor administrative role here at PG--these are my personal views. As such, I have two suggestions regarding the 2015 challenge:

1) We still are not running a Cullen challenge? I have suggested this multiple times, but to my memory I do not recall ever running one. We have run multiple Woodall challenges, and now that project is significantly farther along than Cullen. I would suggest that we replace the Woodall challenge in May with a Cullen one instead.


2) The scoring of the GFN long challenge should be reduced. The bonus for this challenge in 2014 is much too high and has created a situation where those who did not run that challenge suffer on the overall ranking deficit more than others who have not participated in multiple other challenges. For example, only 4 individuals that did not participate in the GFN long challenge this year are in the top 50 overall participants (and all 4 of these have scored in all the other challenges). On the other hand, 6 individuals that did the GFN long challenge, but who have skipped other challenges (often multiple other challenges) are in the top 50. Additionally, another three users who scored in the GFN long challenge, but who failed to score in one or more of the other challenges are also in the top 50.

The GFN WR work is the most exclusive of all our projects (i.e., it excludes all/most CPUs and includes only higher-end GPUs and some mid-tier GPUs), which means that it excludes more users than any other work. Compounding that with an inordinately high bonus challenge credit seems ill-advised. For me personally, it has made this year's challenge series not fun.


____________
141941*2^4299438-1 is prime!

I'm surprised to see no challenge celebrating PrimeGrid's 10th Birthday on June 12. As Woodall Prime Search is the oldest subproject that is still running, I suggest rescheduling the WOO LLR challenge to start or end on June 12, renaming it "PrimeGrid's 10th Birthday", and maybe extending it to 10 days.

The ESP Sieve challenge could then fill the World Expo slot from May 1 to May 4.

Another point: The Lunar Eclipse is on April 4, so date and theme don't fit together for that challenge.
____________

Scott,

TL;DR: I don't mind lowering the point advantage, but it shouldn't be eliminated.

My take on the GFN-WR bonus is a little different. GFN is a project where you have both CPU and GPU participants, but the GPU program is very efficient and runs MUCH faster. Bonus or no bonus, the people with a bunch of big GPUs are going to dominate the leaderboards Even a crazy fast optimized CPU is still going to be 5 or 6 times slower than even a mid-range GPU. For someone such as yourself with a massive amount of computers at their disposal, but with (presumably) comparatively fewer big GPUs, I could see how that could skew the results against you and make it "less fun" as you say.

The bonus for the WR tasks doesn't change that. What it does change whether those GPUs are running the long or short tasks in the challenge. Since there's always an inherent advantage to running short tasks, and we want people running the long tasks, we either need to provide some incentive during the challenge for people to run n=22 rather than n=20, or we need to only run the n=22 challenge. In theory, we could just run an n=20 challenge, but it's n=22 that we want to promote.

That being said, while I think we need an n=22 incentive, I suspect even a small incentive would be effective. Perhaps lowering the challenge point bonus is warranted.

But even eliminating the bonus completely for the n=22 challenge won't change what you observed. Even with equal scoring, there's only 2 15-day challenges this year and one of them is GPU-only. People who don't participate in either challenge are going to be at a big disadvantage. Those two GFN challenges account for 30 challenge days in 2015, while the rest of the year has 36 challenge days. Each of the GFN challenges is therefore almost 25% of the total challenge for 2015. That's a huge chunk of points when you're looking at the overall standings.

Also, I'm a bit concerned that if we lower the bonus too much, all the GPUs will move to n=20. Not only is that not what we want for the project, but it would mean that the GPUs would dominate the n=20 challenge and people with only CPUs are going to be negatively affected.

(FYI Charley makes the decisions regarding the challenge. I'm providing the input on PrimeGrid's needs. The only time I would overrule him is if technically we can't do what he wants.)
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

We have actually run a Cullen challenge this year in May, so that's why thought it'd be nice to have a Woodall again this year.

Scott,

TL;DR: I don't mind lowering the point advantage, but it shouldn't be eliminated...

I suggest running challenges that generate more heat in the northern hemisphere winter. What is the number of active PG members North vs Sourth of the tropics?

I suggest running challenges that generate more heat in the northern hemisphere winter. What is the number of active PG members North vs Sourth of the tropics?

I'm also for Cullen rather than Woodall.

May challenge is overlappind with BOINC Pentathlon starting on 2015/05/05.
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

What if, instead of having simultaneous GFN-short and GFN-WR challenges, we ran a single challenge that accepted results from both projects?

On the plus side, there's a built-in incentive to run the long tasks because of the credit bonus (which is reflected in challenge scores).

On the minus side, people without GPUs can compete -- but almost certainly will not do very well going up against GPUs. It's the "don't bring a knife to a gunfight" scenario.

I'd have to re-write several scripts to make that happen, but it's probably doable.

Along a similar vein, addressing SoB, we could run a "Sierpinski Problem" challenge, where SoB, PSP, and ESP could all be used. Running the longer tasks, which have a bigger credit bonus, would yield more challenge points, so there would be an incentive to crunch SoB, while people with slower computers would still be able to participate.

It doesn't even have to be a 15 day challenge. People with really fast computers can crunch SoB in 4 days. Even my 32 bit Sempron can crunch ESP in 4 days.

Although I'll have to rewrite some of the code, assuming there's no unexpected roadblocks this has the advantage of not messing up BOINCstats.


____________
My lucky number is 75898⁵²⁴²⁸⁸+1

What if, instead of having simultaneous GFN-short and GFN-WR challenges, we ran a single challenge that accepted results from both projects?

On the plus side, there's a built-in incentive to run the long tasks because of the credit bonus (which is reflected in challenge scores).

On the minus side, people without GPUs can compete -- but almost certainly will not do very well going up against GPUs. It's the "don't bring a knife to a gunfight" scenario.

I'd have to re-write several scripts to make that happen, but it's probably doable.

Along a similar vein, addressing SoB, we could run a "Sierpinski Problem" challenge, where SoB, PSP, and ESP could all be used. Running the longer tasks, which have a bigger credit bonus, would yield more challenge points, so there would be an incentive to crunch SoB, while people with slower computers would still be able to participate.

It doesn't even have to be a 15 day challenge. People with really fast computers can crunch SoB in 4 days. Even my 32 bit Sempron can crunch ESP in 4 days.

Although I'll have to rewrite some of the code, assuming there's no unexpected roadblocks this has the advantage of not messing up BOINCstats.

Random thoughts:

Flip the SR5 and SGS challenges, and make the SGS challenge a 1-day affair on March 20. I like the 1-day challenge idea as a test of my boinc management skills.

No SoB challenge == good

I liked the parallel GFN challenges in January 2014 and am happy to see them proposed for next year.

Lunar eclipse maximum is April 14 12:01 UTC.

Since we had "World Animal Day" challenge this year, how about "Octoberfest!" for 2015? (just a name change)

--Gary

p.s. The March 20 challenge could also be styled "Svalbard" or "North Pole" as the eclipse will actually be visible there. OK, in practical terms it probably won't be, as it will most likely be cloudy. And yes I've been to both places, so I'm biased :-)

Since there is code work to do, how about trying to schedule for 2016, since you want to make sure it works before giving it to the public.

n=21 won't start until we're done with n=20.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

n=21 won't start until we're done with n=20.

Only when the n=20 search reaches b=∞ can the n=21 search begin? /JeppeSN

n=21 won't start until we're done with n=20.

Only when the n=20 search reaches b=∞ can the n=21 search begin? /JeppeSN

Considerations
*Do we allow people to submit results from multiple subprojects?

Considerations
*Do we allow people to submit results from multiple subprojects?

Although not impossible, it would make the challenge scoring scripts more complex, and likely make the database queries more complicated as well -- and they're already rather beastly. I'd rather not deal with having to figure out whether the user is running multiple subprojects in the challenge, so my very strong preference would be to allow it. That way I don't need to try and figure out whether a user/host is running more than one project.

Multiproject challenges
The idea has been raised to do a challenge across multiple subprojects on one leaderboard. ...

Number of challenges/total length
I've got signals that this years 12 challenges/60 days was a bit much and I tend to agree.

Even the last one got 1209 challenge points (I repeat: for a single WU).

Thanks Charley, I like the idea of MCP.

Just one comment - putting CUL in with the GFN challenge seems a bit odd to me. The other MCP has PSP,SOB,ESP which are all linked by the fact that they are Sierpinski conjecture projects, and use LLR (with WUs of varying size). I think it might be better to replace the ESP challenge in September with CUL. 5 days is enough for CUL (the WUs are shorter than WOO), and anyway ESP is already included in the January MCP. Plus the CUL tasks should run a little cooler than ESP since the FFT length is larger, which is an important consideration for those who will still be in a warm(ish) climate during September.

Cheers

- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

ESP-LLR is in there twice. How about replacing it in September with either 321 or Woodall?

To match the MPC the first challenge of the year has been reduced from 15 to 8 days, which should also hopefully reduce the cleanup time.

W/r/t the SoB cleanup, how about if after one single normal (60) length period starting on the 18th, re-issue the remaining challenge tasks only to hosts that completed one in the original challenge period or in a later period in the cleanup (mid-March) but took no longer to do it than the challenge itself lasted (15 days). I give one full-60 day window after challenge end to give the world a chance to do cleanup.

I have no idea how difficult this may be from a programming standpoint, but it would at least knock out the much slower hosts, and get the "disappeared" hosts out of the way to. I can see that hosts "trusted" with the shorter window early on may still disappear later, but they'd recycled more quickly to remaining hosts.

Any other machine that shows up after the first 60-day period could still of course run SoB, just not challenge cleanup.

If too difficult, or I've missed something obvious, then never mind. I believe I saw a long time ago that once a WU is created, it's deadline can't be changed, so it might have to be done by causing the server to issue triplicate, "premature" work, if possible.

Cheers, and thanks for the work on the schedule, Charley.
--Gary

W/r/t the SoB cleanup, how about if after one single normal (60) length period starting on the 18th, re-issue the remaining challenge tasks only to hosts that completed one in the original challenge period or in a later period in the cleanup (mid-March) but took no longer to do it than the challenge itself lasted (15 days). I give one full-60 day window after challenge end to give the world a chance to do cleanup.

So looking at the feedback I've done some final tweaking. ESP LLR in September has changed to Cullen LLR and the April MPC has lost its Cullen LLR element, but is staying put. So the final version is:

Date start Date end Duration Project Challenge 1-3-2015 1-11-2015 8 days SoB Year of the Sheep PSP Year of the Sheep ESP Year of the Sheep 3-15-2015 3-20-2015 5 days SR5 Solar Eclipse 4-23-2015 5-1-2015 8 days GFN Long World Expo GFN Short World Expo 6-12-2015 6-15-2015 3 days ESP Sieve PrimeGrid Birtday Challenge 8-10-2015 8-12-2015 2 days TRP Sieve Perseid shower 9-10-2015 9-15-2015 5 days CUL LLR Reign Record 10-4-2015 10-7-2015 3 days TRP LLR World animal day 11-12-2015 11-17-2015 5 days SGS LLR Leonids 12-19-2015 12-22-2015 3 days PPS Sieve Winter Solstice

Do any of those start at a time OTHER than 18:00:00? For now, I've got them all entered at the standard starting time.

The solstice challenge(s) have traditionally started/ended at a time of day to coincide with the actual astronomical event. Maybe the eclipse ones too. But that would be up to you guys, of course.

Question about the scoring: for a multi-project challenge, will there be 3 (or 2) separate leaderboards and scoring, or a single combined leaderboard? This year in January we had essentially two separate challenges running in parallel. Either way is OK with me; just wanted to clarify.

--Gary
____________
"I am he as you are he as you are me and we are all together"

87*2^3496188+1 is prime! (1052460 digits)
4 is not prime! (1 digit)

One leaderboard.

____________
My lucky number is 75898⁵²⁴²⁸⁸+1

What if, instead of having simultaneous GFN-short and GFN-WR challenges, we ran a single challenge that accepted results from both projects?

...

I'd have to re-write several scripts to make that happen, but it's probably doable.

I know the 2015 challenge series is set. But it seems we are in a new era of combined sub project challenges. So why not make the Cullen challenge a combined Cullen/Woodall challenge. They are related and about the same size. This way people have a choice and a slight amount of extra strategy may be used, seeing the projects are about 5-10% different in size/time.
____________
Largest Primes to Date:
As Double Checker: SR5 109208*5^1816285+1 Dgts-1,269,534
As Initial Finder: SR5 243944*5^1258576-1 Dgts-879,713

I know the 2015 challenge series is set. But it seems we are in a new era of combined sub project challenges. So why not make the Cullen challenge a combined Cullen/Woodall challenge. They are related and about the same size. This way people have a choice and a slight amount of extra strategy may be used, seeing the projects are about 5-10% different in size/time.

I know the 2015 challenge series is set. But it seems we are in a new era of combined sub project challenges. So why not make the Cullen challenge a combined Cullen/Woodall challenge. They are related and about the same size. This way people have a choice and a slight amount of extra strategy may be used, seeing the projects are about 5-10% different in size/time.

Mike, or anyone else, have thoughts about this? I would really like to see this happen.

I know the 2015 challenge series is set. But it seems we are in a new era of combined sub project challenges. So why not make the Cullen challenge a combined Cullen/Woodall challenge. They are related and about the same size. This way people have a choice and a slight amount of extra strategy may be used, seeing the projects are about 5-10% different in size/time.

Mike, or anyone else, have thoughts about this? I would really like to see this happen.

Bumping this thread. Anyone else have an opinion on combining Cullen & Woodall for the Reign Record Challenge in September?

I guess Charley is the one to make the final decision, so, Charley where are you.
____________
Largest Primes to Date:
As Double Checker: SR5 109208*5^1816285+1 Dgts-1,269,534
As Initial Finder: SR5 243944*5^1258576-1 Dgts-879,713

Who says old dogs can't learn new tricks?

Looking at the run times of some really slow computers, we've decided to swap the durations of the World Animal Day challenge and the Leonids Shower challenge.

The TRP-LLR World Animal Day challenge is now 5 days. The SGS-LLR Leonids Shower challenge is now 3 days, and will start two days later on November 14th.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1