stats questions

NeoGen · 2005-12-27, 14:16

Hello

I'm trying to see how the teams are faring against eachothers on the long run, so I went to here:
http://www.nfsnet.org/status.html
and downloaded the values on each teams stats page of each project onto excel.
(strangely 3^491+1 appears twice.)

Anyways, I haven't summed up each team's values, but I already got a few questions.
1) Are/were there more of these tables? Like for example, if due to space constrains the oldest ones had to be removed from the site. (If yes, where/how could I get them?)

2) What values should I use for comparing teams against each others? The Workunits or the Lines?

3) Why is it that in the overview section of each of the projects, they never get to be 100% completed? (some over 100%, others below...)

4) Even though those past projects don't show 100% on the overview, they won't be changed anymore, will they? (If not I can just make the sums once and take the results, not needing to get the tables again regularly for updates.)

Wacky · 2005-12-27, 15:57

The values go back a couple of years. I believe that Don has all of the data.

Workunits are the standard to use. That unit of measure is designed to be invariant across projects whereas, even within one project, some "lines" are a lot more work than others. This is because we do more sieving for the small line numbers where the effort is more likely to produce relations. We try to "reward" the participant on the basis of the effort rather than on its productivity since the participant has no control over which assignments he gets (and someone needs to do the less productive parts -- no skimming of the cream)

We don't really KNOW how much work we need to do on a given project until we have done "enough". The percentage complete is just an (educated) guess.

The numbers for older projects stop changing once we have managed to create a usable matrix from the relations collected. Generally, there will be some overlap as we start up a new project. Man of the sievers move on to the next project while a few, principally my personal ones, are "mopping up". In the case of 2,761-, I had to go back and do quite a bit more sieving. Rather than disrupt the many, the effort strung out longer than usual. I would say that it is generally safe to assume that the numbers are "final" when they are near 100% and stop changing for a week.

We would welcome a contribution to our website that summarizes past work and presents it in an understandable manner. Volunteers welcome.

NeoGen · 2005-12-27, 17:23

Thanks Wacky :)

I was wondering why the percentages didn't hit the 100% mark. I guess number field sieving isn't something so linear like "Start here, finish there".

I'm passing through the teams lists and some of them have slight "variants". I guess some people misspelled the team name when registering in the client, and that made the stats server create a new team with the slightly different name. I'll gather the "variants" points and add them to their corresponding main team.

And I guess that it's safe to assume for all these stats tables that I picked up from that page, that they're all finished and done with, right? At least the last piece of news dates from April 2005, so the most recent table must probably be from around that time?

NeoGen · 2005-12-28, 13:50

Well... I made a small vbscript on excel to make the sums from the values through out the tables, and got these results. They are the sums of all the tables that I picked up on that page mentioned above.
How does this look like?
(apart from badly structured, I can't find a way to align things in forum posts)

EDIT: Scratch that last line, I just found out how to do it. Much prettier now!

Code:

1	independent_users				120,598.32
2	_c						12,589.40
3	ugent						6,481.59
4	BigGreen					5,916.91
5	Dataplex					4,251.28
6	The_Ocean_State_Free_Net__osfn_org_		2,402.71
7	US_Distributed					2,150.71
8	TeamPrimeRib					2,011.31
9	Prof__James_Challis__Most_Excellent_UK_Team	1,125.73
10	FreeDC						1,046.27
11	Dutch_Power_Cows				1,024.59
12	AMD_Users					515.60
13	NFSNET_Admin					320.46
14	SwissTeam					227.19
15	Virginia_Tech					205.62
16	Team_Ninja					204.06
17	elm						185.06
18	Russia						183.57
19	Rechenkraft_net					183.52
20	Team_TNT_net					147.43
21	cc1580						98.33
22	CSH						86.24
23	CampusHafnia					80.48
24	2_1137P						67.17
25	Ukraine						65.03
26	Picard						54.43
27	Leo_Viper					28.56
28	VT_Physics					25.68
29	NFSNET_Italia___Team_DPRGI			17.33
30	DesertLabs					16.95
31	Spiderman					14.76
32	wer						12.62
33	Russian_GIMPS_Team				11.89
34	www_hibu_no					8.01
35	Belarus						6.36
36	TeamAnandtech					5.83
37	Israel						3.89
38	FreeBSD						3.85
39	Sieber						3.68
40	france						2.66
41	lhasa						2.59
42	khvSibGuti					2.37
43	praesidi					1.50
44	IISc						1.41
45	Brasil						0.96
46	Jaguar						0.84
47	urbino						0.53
48	rulaz						0.53
49	Predator					0.35
50	massimilianorm					0.32
51	RieselSieve_com					0.25
52	russian						0.23
53	seda						0.11
54	1						0.07
55	0						0.02
56	ted163						0.01
57	123456						0.01
58	Kerkmann					0.01
59	10_223P						0.01

xilman · 2005-12-31, 12:40

Quote:

Originally Posted by NeoGen

Well... I made a small vbscript on excel to make the sums from the values through out the tables, and got these results. They are the sums of all the tables that I picked up on that page mentioned above.
How does this look like?
(apart from badly structured, I can't find a way to align things in forum posts)

Nice work.

I don't know if you'd like to do the same for individual users, or even whether it's worth doing at all.

Paul

garo · 2005-12-31, 15:26

Great work NeoGen. Once these stats can be hosted somewhere, it will likely attract more participants to NSFnet. Not sure if that is a desired outcome

NeoGen · 2006-01-01, 16:06

Thanks guys.
I was recently making a few experiments with the users points within each team. It was working out allright, and I could see who were the top members of my team.

But I also discovered that our fellow Mystwalker is also into it, and much more "professionally" than me! Check out his thread here:
http://www.mersenneforum.org/showthread.php?t=5173

I've volunteered to help him in whatever I can (given my little knowledge of these things). If anyone out there wants to help too, get in contact with him. I'm sure that even just moral support will be appreciated.

2005-12-27, 14:16	#1
NeoGen Dec 2005 2×3³ Posts	stats questions Hello I'm trying to see how the teams are faring against eachothers on the long run, so I went to here: http://www.nfsnet.org/status.html and downloaded the values on each teams stats page of each project onto excel. (strangely 3^491+1 appears twice.) Anyways, I haven't summed up each team's values, but I already got a few questions. 1) Are/were there more of these tables? Like for example, if due to space constrains the oldest ones had to be removed from the site. (If yes, where/how could I get them?) 2) What values should I use for comparing teams against each others? The Workunits or the Lines? 3) Why is it that in the overview section of each of the projects, they never get to be 100% completed? (some over 100%, others below...) 4) Even though those past projects don't show 100% on the overview, they won't be changed anymore, will they? (If not I can just make the sums once and take the results, not needing to get the tables again regularly for updates.)

2005-12-27, 17:23	#3
NeoGen Dec 2005 2·3³ Posts	Thanks Wacky :) I was wondering why the percentages didn't hit the 100% mark. I guess number field sieving isn't something so linear like "Start here, finish there". I'm passing through the teams lists and some of them have slight "variants". I guess some people misspelled the team name when registering in the client, and that made the stats server create a new team with the slightly different name. I'll gather the "variants" points and add them to their corresponding main team. And I guess that it's safe to assume for all these stats tables that I picked up from that page, that they're all finished and done with, right? At least the last piece of news dates from April 2005, so the most recent table must probably be from around that time? Last fiddled with by NeoGen on 2005-12-27 at 17:23

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Overall Stats	Fred	PrimeNet	7	2016-02-10 14:48
CSVs for stats available + New combined stats	opyrt	Prime Sierpinski Project	3	2010-05-31 08:13
Stats	em99010pepe	Sierpinski/Riesel Base 5	66	2008-02-06 19:42
stats on 15k	jocelynl	15k Search	5	2004-01-13 15:45
P4 On Stats	HiddenWarrior	Hardware	2	2003-08-13 14:39

2005-12-27, 15:57	#2
Wacky Jun 2003 The Texas Hill Country 3²×11² Posts	The values go back a couple of years. I believe that Don has all of the data. Workunits are the standard to use. That unit of measure is designed to be invariant across projects whereas, even within one project, some "lines" are a lot more work than others. This is because we do more sieving for the small line numbers where the effort is more likely to produce relations. We try to "reward" the participant on the basis of the effort rather than on its productivity since the participant has no control over which assignments he gets (and someone needs to do the less productive parts -- no skimming of the cream) We don't really KNOW how much work we need to do on a given project until we have done "enough". The percentage complete is just an (educated) guess. The numbers for older projects stop changing once we have managed to create a usable matrix from the relations collected. Generally, there will be some overlap as we start up a new project. Man of the sievers move on to the next project while a few, principally my personal ones, are "mopping up". In the case of 2,761-, I had to go back and do quite a bit more sieving. Rather than disrupt the many, the effort strung out longer than usual. I would say that it is generally safe to assume that the numbers are "final" when they are near 100% and stop changing for a week. We would welcome a contribution to our website that summarizes past work and presents it in an understandable manner. Volunteers welcome.

2005-12-31, 15:26	#6
garo Aug 2002 Termonfeckin, IE 5314₈ Posts	Great work NeoGen. Once these stats can be hosted somewhere, it will likely attract more participants to NSFnet. Not sure if that is a desired outcome

2006-01-01, 16:06	#7
NeoGen Dec 2005 2·3³ Posts	Thanks guys. I was recently making a few experiments with the users points within each team. It was working out allright, and I could see who were the top members of my team. But I also discovered that our fellow Mystwalker is also into it, and much more "professionally" than me! Check out his thread here: http://www.mersenneforum.org/showthread.php?t=5173 I've volunteered to help him in whatever I can (given my little knowledge of these things). If anyone out there wants to help too, get in contact with him. I'm sure that even just moral support will be appreciated.