The "100 mile high" Club!
 

We have been thinking about this for a while, so here it goes!

At the bottom of your [URL="http://mersenne.org/account/"]GIMPS data page[/URL], after you log in, there are some pretty charts. One set shows your work for the last year and the other set shows your "Lifetime Stats".

In particular, in the "Lifetime Stats" section, to the right, there is a "Percentile Rank" chart.

Here comes the fun:

If you have 100% for all six bars (TF, P-1, LL, LL-D, ECM and ECM-F) then send us a screenshot of it and we will mark your account in a special way, somewhat similar to the "Location" field, to indicate that you have achieved the necessary requirements to be a member of the "100%" club!

Now, this metric will be checked (casually) every once in a while to make sure you are, indeed, still a member of the "100%" club. If you find that you have fallen below the standard you are welcome to PM us and let us know, so we can (hopefully temporarily) remove the special recognition attached to your account.

Since the PM system does not allow pictures to be attached, send your submissions to [EMAIL="xyzzy@mersenneforum.org"]xyzzy@mersenneforum.org[/EMAIL] and we will take care of the rest. Please indicate in your message what your forum alias is.

Attached is a sample of what we are looking for. (Well, obviously the attachment does not show 100% for the six bars, but this is the section of the screen we need you to send to us.)

We prefer .png images but we will accept whatever you have.

Note: It is incredibly easy to modify the charts in the page source and there is no easy way for us to determine if you have done so, so we will trust that the data sent is legit.

Your data will also be posted in this thread, to honor your achievements and to provide inspiration to others!

If there is interest, we could also have a "90%" club or whatever else you all want.

Thanks!

That list will be really short... People doing all type of activities are rare, and generally the best LL-ers have no ECM/ECM-F stats, and are not members of the forum (LL is by far the heaviest to reach the top percentile). My advice is to lower the standard, make it 99, 98, even 95, so you can have at least SOME candidates for that list. Except in the case when you already know the person, and this topic has an already known target :D

I could bring my ECM/ECM-F stats to 99% in few days, but to raise the LL would be a long way and many months...
So, how about a 96% club? :razz:
[ATTACH]8491[/ATTACH]

[edit: another idea would be to renounce to one of the bars, or mostly two, for example if a guy has 100% LL, DC, P-1, TF, then he may not be required to be top ECM-er, or other combinations].

Set the bar at 99%; Invite people to become one of the 1%-ers.

Toss some ideas out and we will set up a poll so you all can choose.

The forum has a "star" function that allows us to choose the color and number of stars to display in the user info box next to every post.

We only have 100% in two categories so we are interested in alternative ideas as well!

:mike:

I would have a lot of LL work to do, and I've never done ECM for GIMPS.

[QUOTE=Xyzzy;309644]Toss some ideas out and we will set up a poll so you all can choose.
:mike:[/QUOTE]

Well three other possibilities would be:

1. 100% of at least 3 out of 6 categories.

2. At least 99% on all 6 categories.

3. To be on top 100 list on all categories.

Couldn't we also have a kind of nomination of this years (or half year or three months) greatest effort/find. Which could be anything like facorization to the highest level; The largest factor found; The smoothest k-value; The largest LL. With a kind of nomination procedure and general poll. Something like a Mersenneforums Oscar? :smile: And of course there would be something like a goldstar beside that users name?!!!

I mean we could have an Oscar-nomination thread and a poll about which effort / find was greatest that year / period. And anybody can nominate his (or hers!) greatest find/effort.

6 stars
 

Well, you have 5 or possibly six categories. Designate them by different stars (bees, snails, icosahedrons or other symbols). 100% could be gold, 99% silver, 98% bronze. The positions of stars could be associated with particular category. Absence of a job type (or not enough percentage) could be empty space or other symbol. I wonder how many will have 6 or even 5 golden stars (1 or 2 people), so if you do not allow some deviation from golden-six rule, your club might be very lonely place :smile:.

How to make sense of these ranks? I have done tons of LL work and only an occasional P-1. My P-1 rank is 96%, while LL is 78%.

Is that just relative to other users doing the same type of work?

[QUOTE=TObject;309657]Is that just relative to other users doing the same type of work?[/QUOTE]
Yes. What you're seeing is that a lot more people do a lot more LL work than P-1.

How about just giving us a star for each 100% we have?

So, 100% is anywhere from 99.5% to one hundred. Right?

This is my lifetime

Lifetime for me:

So who is going to take the bull by the horns and volunteer to research, poll, gather information and implement this idea?

[url]https://www.vbulletin.com/docs/html?manualversion=30807603[/url]

The task would, we think, have to be done with usergroups. Very easy stuff.

PM your curriculum vitæ to us and perhaps YOU could be the lucky winner!

:whistle:

Heh. I skimmed the manual, didn't see a way to decouple rankings from post count, then read a bit on user groups and saw that might work, and then thought "I'm glad Xyzzy came up with this; he must have an idea in mind."

My stats, though I haven't done anything in a month, and have otherwise been with reduced throughput since March or April or May or thereabouts.

Here's mine, though I don't have anything near the firepower many here do. Thank goodness for "lifetime"! :smile:

Xyzzy wants that I do more ECM...

Mine:

My record just for these last few minutes only. You would cry if you saw my all time record.[QUOTE=Xyzzy;309634]Note: It is incredibly easy to modify the charts in the page source and there is no easy way for us to determine if you have done so, so we will trust that the data sent is legit.[/QUOTE]I don't have a graphic editor program. I've never heard of Paint or Photoshop and I don't know what they do and I can't imagine what they could be used for.

Father: Son! Come here now.
Son: Dad, I didn't eat all the biscuits.
Father: I never mentioned any biscuits. What have you been up to?
Son: Ermm ...

[QUOTE=retina;309688]I don't have a graphic editor program. I've never heard of Paint or Photoshop and I don't know what they do and I can't imagine what they could be used for.[/QUOTE]
It's nothing that [URL="http://chart.apis.google.com/chart?cht=bvs&chd=t:110,105,137,21&chs=300x140&chxt=x,y,x&chxl=0:|TF|P-1|LL|LL-D|2:|%28110%%29|%28105%%29|%28137%%29|%2821%%29&chbh=30,10&chtt=Percentile+rank"]complicated[/URL].

I have pitiful firepower. No chance for anything for me :no:

4 + 2
 

4 X 90%+
2 X 100%

Personally:

I think 6 X 98% is doable.
6 X 99 a stretch.
6 X 100 .... special club indeed....not one I would likely get membership in.

Well, I never minded about the lower graphic... Now, I'm impressed! :smile:

Luigi

In fact, about ECM/ECMF, everything we should do is to convince more people to join the fray. If more people participate in ECM/ECMF factoring, our percentile will raise even if we do nothing. It is different to be position 7 in a hundred (93%) or be position 7 in a thousand (99%).

For me, and for anyone below position [strike]10 (!!)[/strike] 5 (five!) on ECMF top, this is the only solution to reach 100% percentile for ECMF. If ET_ with that big score and third position in top is just 99%, then we are all hopeless... Without increasing the number of participants, this will be more difficult than LL - [B]almost impossible[/B], no matter your resources, without cheating**. You must get the first/second position on top to have 100% percentile, and did you see the scores of those two guys?

Maybe this is what xyzzy wants, in fact? Increasing the numbers of ECM/ECMF participants? Why? So he can get better percentiles too? :razz: It works for me, however! I will also get better without doing anything! hehe...

[edit ** cheating is easy on ECM/ECMF, the server still has that bug to give you multiple credits if you submit the same curves many times. So you can do 5 curves and get credit for 50 or 500 or more - no joke, I tried it]

A small trick for ECMF stats
 

If you have a big iron and lots of RAM, you can do ECMF on F12-F13-F14 with GMP-ECM using the default B2 range chosen by the program instead of B2=100*B1, submit your results to George and gather lots of points. :smile:

If your RAM is below 8GB, you can still use a B2 greater than the one chosen by Prime95, and get your premium GHz/day.

Luigi

[QUOTE]If you have a big iron and lots of RAM, you can do ECMF on F12-F13-F14 with GMP-ECM using the default B2 range chosen by the program instead of B2=100*B1, submit your results to George and gather lots of points.[/QUOTE]Too bad there isn't a provision to award mega points for finding a Fermat factor. (Even if it was with a different program!)

:razz:

You all are tossing out great ideas. Keep the ideas coming and surely that there will be a consensus about how to manage this!

We still need someone to step up and take control of things. We heard that ET_ was thinking about it…

:et_:

I redid the calculus according with what I said in post #25. If the goal is not to bring more people into ECMF factoring, then you should really exempt the ECMF bar from the 100% requirement. It is almost impossible for anybody except Buckle (first guy in top) to get 100% percentile for ECMF [U]without cheating[/U], for many years from now, regardless of their hardware. You should exempt ECM at all, it looks like an invitation to cheat, which can have bad effect on the number of curves, as long as cheating is possible (see post #25).

[QUOTE=LaurV;309727]I redid the calculus according with what I said in post #25. If the goal is not to bring more people into ECMF factoring, then you should really exempt the ECMF bar from the 100% requirement. It is almost impossible for anybody except Buckle (first guy in top) to get 100% percentile for ECMF [U]without cheating[/U], for many years from now, regardless of their hardware. You should exempt ECM at all, it looks like an invitation to cheat, which can have bad effect on the number of curves, as long as cheating is possible (see post #25).[/QUOTE]

It is trivially easy to cheat on P-1 also. I will leave out the details of how.

[QUOTE=frmky;309692]It's nothing that [URL="http://chart.apis.google.com/chart?cht=bvs&chd=t:110,105,137,21&chs=300x140&chxt=x,y,x&chxl=0:|TF|P-1|LL|LL-D|2:|%28110%%29|%28105%%29|%28137%%29|%2821%%29&chbh=30,10&chtt=Percentile+rank"]complicated[/URL].[/QUOTE]I'm too old-school to understand all this modern computery networky stuff. All I need is a block of wood and some fencing wire and I can make anything.

[QUOTE=LaurV;309727]It is almost impossible for anybody except Buckle (first guy in top) to get 100% percentile for ECMF [U]without cheating[/U] ...[/QUOTE][QUOTE=axn;309728]It is trivially easy to cheat on P-1 also. I will leave out the details of how.[/QUOTE]Why even bother to cheat the backend system when [url=http://mersenneforum.org/showthread.php?p=309688#post309688]the front door is wide open[/url].

Nope, not even close. Unless, of course, someone is willing to buy me a couple dozen of those 40-core HPC servers. :D

My vote....
 

6 possible stars (one for each category)
Platinum for 100%
Gold for 98% (or 99%)
Silver for 95%
Bronze for 90%
(Diamond? Kryptonite? if you have all 6 Platinum)

[QUOTE=petrw1;309713]4 X 90%+
2 X 100%

Personally:

I think 6 X 98% is doable.
6 X 99 a stretch.
6 X 100 .... special club indeed....not one I would likely get membership in.[/QUOTE]

Hmmm...I was already there (90%+) 15 months ago......(365 too; not any more)
[url]http://www.mersenneforum.org/showpost.php?p=261251&postcount=696[/url]

Those are all interesting ideas.

How about "in the top 100" for each category? You get a star for each one?

(Although the 99% club sounded good.)

So I was bored....and I did some ciphering.
 

According to my ciphering (and I could very well have messed up) but this should be close:

There is currently NOONE with 6 100%-stars (or 7 in my stats: I counted 1 for Overall as well).

The closest are (of 119 with at least 1 ... or 104 if you exclude Overall):
The number after each heading is their rank in that list.
[CODE]
[SIZE="2"][FONT="Courier New"]
Member Name All TF P1 LL DC EC EF Stars
Sturle Sunde 9 16 6 10 17 3 - 6
ahmerali 4 25 39 5 6 - - 5
linded 3 46 4 3 3 - - 5
TheJudger 7 5 2 18 2 - - 5
GrunwalderGIMP 17 15 15 - 4 - - 4
Torchwood Institute 26 24 21 - 11 - - 4
unconnected 24 - 25 25 - 5 - 4[/FONT][/SIZE]
[/CODE]

If you have the data already, it would be very interesting to know where the 100%, 99%, 98%, 95%, 90% break in the top. For example, being in the first 100 at DC or P-1 puts you in 99% percentile, but being in 50th at ECM only puts you in 80th% percentile. Then xyzzy can think more about "top100 on each category", versus "top custom number for each category".

If you do this table I promise you I let you overpass me in gpu72 DC top :razz:

[QUOTE=Xyzzy;309826]Those are all interesting ideas.

How about "in the top 100" for each category? You get a star for each one?

(Although the 99% club sounded good.)[/QUOTE]

Even for TF? :razz: I think most people are here.

[QUOTE=petrw1;309829]According to my ciphering (and I could very well have messed up) but this should be close:

There is currently NOONE with 6 100%-stars (or 7 in my stats: I counted 1 for Overall as well).

The closest are (of 119 with at least 1 ... or 104 if you exclude Overall):
The number after each heading is their rank in that list.
[CODE]
[SIZE="2"][FONT="Courier New"]
Member Name All TF P1 LL DC EC EF Stars
GrunwalderGIMP 17 15 15 - 4 - - 4
[/FONT][/SIZE]
[/CODE][/QUOTE]

I think you're missing my #16 ranking in ECM.

Graff (a.k.a GrunwalderGIMP)

[QUOTE=Graff;309866]I think you're missing my #16 ranking in ECM.

Graff (a.k.a GrunwalderGIMP)[/QUOTE]

Ranks are only listed if they are in the 100%.

PS forgot to mention. I excluded ANONYMOUS because I had no way of knowing if it was the same person.

[QUOTE=LaurV;309831]If you have the data already, it would be very interesting to know where the 100%, 99%, 98%, 95%, 90% break in the top. For example, being in the first 100 at DC or P-1 puts you in 99% percentile, but being in 50th at ECM only puts you in 80th% percentile. Then xyzzy can think more about "top100 on each category", versus "top custom number for each category".

If you do this table I promise you I let you overpass me in gpu72 DC top :razz:[/QUOTE]

You mean let me stay ahead. Have you checked lately.

I'll see what I can cipher this weekend.

[QUOTE=Xyzzy;309826]Those are all interesting ideas.

How about "in the top 100" for each category? You get a star for each one?

(Although the 99% club sounded good.)[/QUOTE]

I like the idea of Top-100. It's just cool to have a 2-digit rank.
Though some are very easy (ECM, P1) and some (LL) are much harder.

[QUOTE]I like the idea of Top-100. It's just cool to have a 2-digit rank.[/QUOTE]You sound like you are interested in managing this awesome program. Maybe PM us sometime?

:smile:

[QUOTE=petrw1;309872]You mean let me stay ahead. Have you checked lately[/QUOTE]
[offtopic]
Have you check the fact that I didn't report nothing for a while? I had no net in my house for a week, and I have about 300-400GHzDays accumulated on the pipe to report :smile:. And chalsall "cheated" me of 60GHzDays of DC (because he was afraid I will overtake him, which happened anyhow! :razz:), if you click on my name you will see I have 60 GHD more DC than shown in the top-DC list. The explanation is that I did some "private" DC assignment in the past, and Chris had to add it by hand, but he added to my DC credit in my individual report, and (wrongly) added to LL credit in the top-list. Now as a programmer, I still have no idea why the tables do not use the same source of data, but that is not my business. The God's ways are mysterious. The same discrepancy between my assignments, sometime my individual report show some numbers and the assignment list shows different numbers...
[/offtopic]

Little late to the thread, but I think the bar needs to be set pretty high, I've been doing this for a lot less than a year and I'm above 90% in all categories. (I'm probably also in the 90th percentile of assignments lost to crappy rebuilt Dell server failures.)

[QUOTE=chappy;310472]Little late to the thread, but I think the bar needs to be set pretty high, I've been doing this for a lot less than a year and I'm above 90% in all categories. (I'm probably also in the 90th percentile of assignments lost to crappy rebuilt Dell server failures.)[/QUOTE]

Yeah ... 90's nothing to be ashamed of ... but the 98-100 can take a bit of work; some types more than others.

I know, I know, the clock is ticking....been busy.
 

However, I haven't stopped thinking about this project....
I have a few questions:
1. What is the "widely accepted" formula for calculating percentiles?
2. Do we really want to report percentiles?
2a. Or percent rankings (what George does)?
2b. Or do we want to report Top-X? i.e Top-10, Top-25? Top 100?
2c. Or ???
3. What is the cut off percent? Top-xx (We don't want the reports too big)
4. For Lifetime and YTD?
5. For all 6 categories or work (All, TF, P1, LL, DC, ECM, ECM-F)?
6. What do I use for sample size?
a. Total number of names on the list?
b. Number of names over 0 points? I believe this is to 0 to 3 decimal points.
c. Highest rank listed? I believe this is anyone truly over 0.

Attached is a sample from the Lifetime All report from about a week ago.
Might be kind of tiny....sorry.
Columns are:
-Ordinal Rank in column
-Member Name
-Total GhzDays
-% Rank as George computes it on your Summary (** 6. c.)
-Percentile as I understand it. (** 6. b.)
** In the last 2 columns the number on top is the sample size (ss) from 6. I used.

My formulas:
% rank: (ss-rank+1)/ss. Result is a percent.
Percentile: INT(100-(100*((rank-0.5)/ss)). Result is the integer percentile.

I have an issue.....
 

I thought a better format might be a table like so:
Ordered more or less by #of G/S/B.
Gold might be 99th Percentile
Silver 95th -98th
Bronze 90th - 94th.

HOWEVER....names are NOT unique; muchly NOT so.
I looked at ALL names in the Life-Time all report.
There were over 16,000 entries.
Besides the obvious "ANONYMOUS" that appears about 2300 times; and by definition want to be excluded ...
there are over 300 other names that appear anywhere from 2 to 11 times for a total of 1100 other duplicates.
I didn't verify if any of these are currently in the G/S/B status but I can't rule out that they will be some day.

At the risk of excluding or falsely reporting some members I am at a loss for how to proceed...

Comments? Suggestions? Advice?

Reached 92+% in all 6 categories. I'm happy with that, and I won't get to 100% unless I win the lottery or get a new job where I can run Prime95 on machines.

You won't be able to win 100% for ECM-F even if you win the lottery. Or... well, you have to buy that guy's account and join the user to your one. Or you can create 1000 user accounts and fake some ECM-F reports, give them a curve each, they will finish in few days, this way you lower the watermark, otherwise you need/have to be the first to have 100%, being second you get 99.3 that is rounded to 99.

"Fill in the yellow cells" version:
[ATTACH]8969[/ATTACH]
(first table is what one need for some percentile, second is my current stats, last is the check/probe. So I can afford to lose 7 places on TF, but I have to gain 226 places on LL :razz:)

Actually, #6 in ECM will give 100% :smile:
[ATTACH]8981[/ATTACH]

Hm.. if so, there is a calculus mistake on PrimeNet (up to now, my calculus matched exactly, I did not find any mistake). 6*100/1101 is 0.54xxxx, you need under 0.5 to be rounded in the first percentile, otherwise is rounded as 1. (i.e. you are 99.45xxx which is rounded as 99).

Hey Xyzzy, how about this? [URL="http://www.vbulletin.org/forum/showthread.php?t=196244&page=44"]Yet Another Awards System[/URL]

Not gonna last (it is getting too warm here with the extra space heaters), but 100% in 5 major categories (TF, P-1, LL, LL-D, ECM) :

Gratz!!! Very very nice :smile:




Now you need 100% in ECM-F :wink: There are 324 in the list atm, so 2nd place ~ 99.38% which *might* be rounded up to 100% since it doesn't always round correctly. In that case you "only" need ~ 36000 Ghz days which is ~ 11% more compared to your 308k total so far :big grin:
But if you need #1 you need to beat Buckle at 150k or 50% more than what you got total now :cool:

Yes, the rounding is funky. I should be at 99 in LL and ECM, but I didn't make the rules (I just enjoy them).

ECM-F isn't going to happen. ECM only happened because a few years ago I accidentally set the wrong worktype and completed one ECM exponent, and having that one category at 01% bothered me.

*BUMP*

Reached 97+% in all 12 categories in lifetime stats, 98+% in 11 categories and 99+% in 6 categories.