Anyone interested in a 12 Days of Christmas Challenge on PRPNet? This challenge would rotate between projects, with a different project for each day of the challenge. It would run from Christmas Day through January 5th.

Well, it sounds interesting but aren't some wu's over 1 day long? Especially on somewhat older machines.

GFN524288 is GPU only if you want to do it in one day, unless you have a particularly fast CPU. It's about 50 hours on a Core2.

____________
My lucky number is 75898⁵²⁴²⁸⁸+1

I haven't run every project, but I would guess 5oB is out as it took about 26 hours on my i5 2500k (@stock).

You could do a 12 days challenge where you have to run all of the PRPNet projects an equal amount. People can switch back and forth on their own schedule. It would be sort of like a road rally rather than an all-out race. Getting to the finish line first isn't the only criteria; you also have to get there in the correct time.

Maybe it could work like this:

Total score is based upon two factors, the number of total points earned across all of the projects, minus a penalty for having less than a perfectly even distribution between the projects. The first part is pretty simple. The penalty would work as follows:

1) If there's N projects, each project is expected to contribute exactly 100/N percent of the total points. Of course, no project is going to do exactly the correct amount, so there will be some error for each project. If there's 4 projects, each should contribute 25%. In reality, one might do 45%, another 35%, a third might do 20%, and the fourth might do 0%.

2) To figure out the penalty, compute the "error" for each project as the difference between the actual amount done and the "correct" amount that should have been done. In the example above, the error amounts for those four projects would have been 20%, 10%, 5% and 25%.

3) Use the largest of all the error values. This penalty -- 25% -- is then applied to the total points earned.

So, if you earn a million points, but only crunch one out of four projects, the error for each of the four projects would be 75%, 25%, 25%, and 25%. The largest error is 75%, so you would incur a 75% penalty and only be awarded 250,000 points for the challenge.

This has the advantage of being rather tolerant of individual's personal schedules, since you can start and stop crunching at times of your own choosing. Careful planning will be just as important as raw crunching power. Knowing how long each WU takes on each computer, and how many points you get for each one, will be very important. There's also strategic decisions to be made: Do I run project X that gives me more points per hour, even though I'll get a larger penalty? (Anyone remember what an integral sign looks like???)

(Yes, this is complicated enough to probably be more of a pain than a challenge, and it seemed like more fun when when I started writing this than when I finished. I put it out there as a talking point. Maybe it will inspire a better idea from someone else.)
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

I agree that switching between projects from day to day would be difficult. How about a schedule like this:

Dec 25: PPSElow (open as many gifts as you can)
Dec 26-Jan 4: 27, 121, ESP, SGS, and Mega (any or all)
Jan 5: PPSElow (open remaining gifts)

Limiting to base 2 only limits the number of projects and gives a level playing fields WRT PRPNet scores. There is already a planned GFN challenge, so those projects can hold off. We'll need to get challenges for PRS, FPS, GCW, and SR5 for the future.

Im in, looks like a great idea!


(Personally, I would love a push for 27, even if just a day. Tests are currently 5 hours each and we are getting close-ish so that we can re-merge with 121)

Just a reminder that this starts on Sunday.