It seems that this has gotten lost in the sea of forum activity. With so many polls and contests going on, it's easy to see why! I was wondering if we had ever determined who would run such a poll, and if it would run at all. I, for one, think it would be very interesting to compare the results of an artists poll to both the AM list and the Survivor results, as each represents something different.

So, has anyone agreed to run this poll? If so, when? I believe it should not be revealed until after the results of the Artist Survivor contest, but theoretically, voting could begin at any time.

Also, is there interest in such a poll?

I'm interested. If somebody wanted to set up a spreadsheet for me and a formula I'd be interested in running it, but I think I'd screw it up too much doing that by myself. But, I'd be up for the data entry.

I believe the plan was starting about 1 month before Survivor ended, and I believe you volunteered Moonbeam.

I'm happy to do it, but I wasn't sure if someone else wanted a turn. Any other thoughts?

This poll is a very good idea.

I've been using Henrik's formula to compile my own list, with some very nifty Excel formulas (for me anyway) so that the artist ranking automatically changes when I change my song/album rankings. So I'm ready for this.

I'm psyched for this poll. I say we doubt in between the end of Survivor and the beginning of the Philly.

doubt = do it

I've been wondering if there's some poll I could use to test out my compare-based ranking algorithm on lists instead of just reviews.

The only thing is, I probably need to rewrite the way the data is internally stored, or else entering data would be way more work than it needs to be for me.

so i've started making my list and i have a couple questions for anyone who wants to answer:

how do you guys deal with songwriters like burt bacharach and hoagy carmichael? they've both recorded some versions of their own songs, but they're not the best versions or not the best songs. whereas someone like dionne warwick, who's most closely associated w/ bacharach, has had the opportunity to sing some of the greatest songs ever written, although i don't think she's an outstanding singer.

also, how do you deal w/ artists who operate under several different names? i'm thinking of people like stephin merritt (the magnetic fields, future bible heroes, the 6ths, gothic archies), will oldham (palace, bonnie 'prince' billy) and kool keith (dr. octagon, dr. dooom, etc.) would you enter the artist as stephin merritt or as the magnetic fields?

this applies to rap as well, where's there's lots of guesting. for instance, wu-tang really only has 4 albums, but if you include all the "solo" projects (i put that in quotes because at least half the gang it present on almost every record) you've got quite a body of work.

also, i encounter the same problem with producers like dre and timbaland. dre only has 2 good albums, but i'd like to give him credit for all the other production he's done.

gosh, this is tricky.

Hey Moonbeam, sure I'm interested in doing an artist poll.

My artist line up has been based on albums only, but working on some kind of way to let songs influence as well. And Henrik's formula will do the rest, I guess.

Just one questions: When exactly will we submit it? May, June, later? I'm a bit too busy at the moment, so if it was my say, preferably later towards summer ...

It seems that there is enough interest in such a poll. If nobody is dying to run it, I'm happy to. I think Schleuse has said that he would hold off on running the Philly experiment until voting for this had closed, so I'm hoping to take submissions during the last month of the Survivor game, so that the results can come soon after that.

I would assume everybody makes their own criteria for judging an artist. I would certainly give Brian Eno credit for more than just his solo albums.

Heh, I can't get into many Dr Dre albums. Not just because I can't stand 'gangsta' lyrics, but because the sheer number of times he references his label ticks me off. He's a genius music producer because it's so good at franchising his artists, but franchising has nothing to do with the actual music.

stephan

Reading this poll, I am very intersting by your excel formulas which link artists, songs and albums poll
How did u do ?

Basically I got a formula that checks the tab 'Albums' for the first occurrence of a certain artists name and then takes the corresponding ranknumber and that's the result of the formula.

{=IFERROR(INDEX(Albums!$A$2:$B$251;SMALL(IF(LEFT(Albums!$A$2:$B$251;LEN(B2))=B2;ROW(Albums!$A$2:$B$251)-ROW($B$2)+1;ROW($B$251)+1);1);1); $A$99)}

These formulas only work when you ctrl+shift+enter them, they're some kind of special formulas (I forgot the name). To explain them a little more:

"Albums!$A$2:$B$251" makes it look through the tab Albums from A2 to B251
The "$B$2" you see everywhere is the artist name I entered in that cell (this is to make copying the formulas easier)
"$A$99" refers to the number it'll take if there is no occurrence, which is 350 in my case (100 above the 250th and last spot)

Then another formula for the second occurrence (the only change is the 1,2,1 instead of 1,1,1 near the end) etc. until the 6th:

{=IFERROR(INDEX(Albums!$A$2:$B$251;SMALL(IF(LEFT(Albums!$A$2:$B$251;LEN(B2))=B2;ROW(Albums!$A$2:$B$251)-ROW($B$2)+1;ROW($B$251)+1);2);1); $A$99)}

Then I did the same for Songs until the 6th occurrence with slightly different numbers (list is longer):

{=IFERROR(INDEX(Songs!$A$2:$B$500;SMALL(IF(LEFT(Songs!$A$2:$B$500;LEN(B2))=B2;ROW(Songs!$A$2:$B$500)-ROW($B$2)+1;ROW($B$501)+1);1);1); $A$100)}

All those results are compiled by Henrik's artist formula:
=((LN(LN(10+C2))+LN(LN(10+D2))+LN(LN(10+E2))+(LN(LN(10+F2))+LN(LN(10+G2))+LN(LN(10+H2)))))+(0.6*(LN(LN(10+J2))+LN(LN(10+K2))+LN(LN(10+L2))+LN(LN(10+M2))+LN(LN(10+N2))+LN(LN(10+O2))))

And the result of that is my artists score. You will have to pick the names of the artists yourself, but you could just enter 200 and see which ones come out on top. It takes only the first two words in the name of an artist into account, so if you have "Neil Young" and "Neil Young and Crazy Horse" it'll both count them as Neil Young.

I hope that all makes sense. If you want I can upload an example file.

yeas, I think I'd like it...

I've just used my own lists as an example (I deleted all the frills like releasedates, ratings, etc.) because making a new example file is a lot of work. This should explain it fine too though. Click me!

The ranking of the artists is a little off because I cheat a little and give certain artists penalties for having very few albums or just when I think they're ranked too high. The formula is a little partial to very high spots rather than multiple spots. (Kim Carnes will end up on the 20th spot for example, by having just 1 album and 2 songs in the lists, so I gave her a penalty and she's now 39th. Still a little high I suppose, but oh well)

thank you very much !
I'll have a look at it

It's great to see this thread resuscitated! I actually was going to do it myself to pose a question.

I don't quite know how to resolve questions of the ilk of "What makes an artist a separate entity?" I'm thinking to use AM as the guide. So Prince and the Revolution gets lumped in with Prince, and Parliament/Funkadelic are treated like one act, although some may question the veracity of that. There are a few other borderline cases like Gary Numan/Tubeway Army, but rather than get caught up in minutiae, I'm proposing to use the AM Artists list as the deciding factor, unless there is some reason I'm failing to consider that would suggest a different guideline. I figure that most of the artists nominated for the poll will be recognized by AM somehow, and for those borderline cases not represented by AM, such hair-splitting probably won't change the final results regardless of how such artists are treated.

And Stephan, I'm definitely going to take a look at your formula and see how it may impact my artist ranking!

I haven't changed Henrik's formula for calculating the artist score, my formulas just make it a little easier to see the effect of changes in my album/song lists to my artists list.

The formula definitely opened my eyes to a few of my ranking disparities!

I also thought I should open up a question. What should the maximum number of artists per ballot be? 100? 200?

I guess that 100 would be enough (and less work for you)

I don't mind the work if 200 seems like a better maximum. In fact, I think there will be much less work involved with this poll, as I believe that there will be much more overlap than in the song poll.

Yes, but still i guess 100 will be anough for me

Thanks, Stephan, for the formula and Excel example! At the moment, I have a ranked list of 1000 albums and 200 songs, still some work to do on the song side there ...

I agree with nicolas, 100 artists should do the trick for me as well. Beyond that limit, the artists are not 'great' anymore and as mentioned above, I haven't ranked enough albums and songs yet to get a meaningful list above 100.

I'm personally gonna stick to 50.

I've got a list of 130, but I'll narrow it down to 100. I'll probably do what I did with the song poll, and put it together very quickly when it's time to post it.

(I know it's ironic that the guy who runs lots of games involving numbers prefers to put his lists together by feel and gut.)

What weights are you going to use? In my opinion, higher rankings should be weighted heavier in this game just because we're not going to see as much differential.

Hi John-

This comment slipped by me! I wasn't planning on changing the point system, although I suppose we could have a discussion about it. I'm hoping to open up voting lines soon.

I think it would be a good idea to give more points to higher rankings. In the albums contests and especially the songs contest there wasn't that big of a difference between say, #10 and #95. But in the artists poll there will be an astronomical difference between #10 and #25.

What about a formula like: Points = 100/(Position^0.4)?

Here are the point values for positions 1-100:

Position Points
1 100.00
2 75.79
3 64.44
4 57.43
5 52.53
6 48.84
7 45.92
8 43.53
9 41.52
10 39.81
11 38.32
12 37.01
13 35.84
14 34.80
15 33.85
16 32.99
17 32.20
18 31.47
19 30.80
20 30.17
21 29.59
22 29.04
23 28.53
24 28.05
25 27.59
26 27.17
27 26.76
28 26.37
29 26.00
30 25.65
31 25.32
32 25.00
33 24.69
34 24.40
35 24.12
36 23.85
37 23.59
38 23.34
39 23.10
40 22.87
41 22.64
42 22.42
43 22.21
44 22.01
45 21.81
46 21.62
47 21.44
48 21.26
49 21.08
50 20.91
51 20.75
52 20.59
53 20.43
54 20.28
55 20.13
56 19.99
57 19.84
58 19.71
59 19.57
60 19.44
61 19.31
62 19.19
63 19.07
64 18.95
65 18.83
66 18.71
67 18.60
68 18.49
69 18.38
70 18.28
71 18.18
72 18.07
73 17.98
74 17.88
75 17.78
76 17.69
77 17.60
78 17.51
79 17.42
80 17.33
81 17.24
82 17.16
83 17.08
84 16.99
85 16.91
86 16.83
87 16.76
88 16.68
89 16.61
90 16.53
91 16.46
92 16.39
93 16.32
94 16.25
95 16.18
96 16.11
97 16.04
98 15.98
99 15.91
100 15.85

I still don't think that's enough of a differential but maybe that's just me. I'd be kidding myself if I didn't put Bruce Springsteen somewhere in the 75-100 range. But, I wouldn't want to give him only 70 points less than I give a band like The Zombies or The Beach Boys or whatever. The difference between the #20 artist and #60 artist is enormous. I'd almost want the #1 artist to get 1000 points and each number after that loses 50 points or whatever makes it work out. Artist lists are much different than songs and album lists. Just my opinion.

Scaling it to 1000 wouldn't really change anything- it's relative proportions that matter. A linear scale is probably best from that angle, but then the top rankings are so close. Any other suggestions from anyone?

Perhaps a combination of a linear and a non-linear scale? For example

Points = [100/sqrt(position)] + [101-position] / 2

1 100
2 84.86
3 77.87
4 73.5
5 70.36
6 67.91
7 65.9
8 64.18
9 62.67
10 61.31
11 60.08
12 58.93
13 57.87
14 56.86
15 55.91
16 55
17 54.13
18 53.29
19 52.47
20 51.68
21 50.91
22 50.16
23 49.43
24 48.71
25 48
26 47.31
27 46.62
28 45.95
29 45.28
30 44.63
31 43.98
32 43.34
33 42.7
34 42.07
35 41.45
36 40.83
37 40.22
38 39.61
39 39.01
40 38.41
41 37.81
42 37.22
43 36.62
44 36.04
45 35.45
46 34.87
47 34.29
48 33.72
49 33.14
50 32.57
51 32
52 31.43
53 30.87
54 30.3
55 29.74
56 29.18
57 28.62
58 28.07
59 27.51
60 26.95
61 26.4
62 25.85
63 25.3
64 24.75
65 24.2
66 23.65
67 23.11
68 22.56
69 22.02
70 21.48
71 20.93
72 20.39
73 19.85
74 19.31
75 18.77
76 18.24
77 17.7
78 17.16
79 16.63
80 16.09
81 15.56
82 15.02
83 14.49
84 13.96
85 13.42
86 12.89
87 12.36
88 11.83
89 11.3
90 10.77
91 10.24
92 9.71
93 9.18
94 8.66
95 8.13
96 7.6
97 7.08
98 6.55
99 6.03
100 5.5

I understand, that DrDre's formula does not work very well here for the artists poll. BTW, haven't heard from him for a while, seemed such a nice poster

For a moment, I was thinking to suggest the option, to do it as in the Pazz & Jop poll (e.g. max. 100 points per artist, min. 50 artists, max. 100 artists, max. 5000 points to assign) but that might be too much trouble for both the voter and moreover for Moonbeam ...

And then saw your formula, Henrik, which looks pretty good.

It's a tough call to pick a formula. I guess we have to consider what we anticipate the case might be with the diversity of the lists, and what we want to rate highly, such as whether we value a few high placements in lists more than lots of middling to low placements in lists, etc.

I don't think that the diversity is going to be very high. We'll definitely see some outliers but the top 100 AM artists are going be well represented.

What if we did it like the 2007 album list? You get 300 points or something like that and put a cap on the max points you can give one artist.

Though I'd bump this and see if we have a deadline to submit.

Hi Slush-

I envision a deadline of the end of July, but I want to figure out the point formula before taking ballots, in case that has some impact (although I don't think it would in most cases) on the lists.

We really need to iron out a point system soon! Are there any other potential formulas to be considered, or should be choose between the hyperbolic formula and the linear/hyperbolic mixed formula?