I think mathematical formulae apply to music as well. For instance, N = the ratio of good songs/mediocre songs for any given band or artist.

It's my opinion that in the history of rock n roll, Led Zeppelin has the highest N ever.

Which is not to say that Zep is the greatest band ever, that is a subjective viewpoint you may agree or disagree with.

I submit however that N is an objective measure, and that of all bands, Led Zeppelin has the highest N.

There is some an element of subjectivity, I suppose, in that it does come down to how many of a band's songs you consider to be good and how many you consider to be mediocre.

So my gut says you need to bring in some kind of bell curve so truly great songs cancel out duds.

The example I'd give is Elvis Costello. Easily one of the all-time great rock songwriters, but the downside of his prolific career is that there were a few mediocre moments — but the sheer number of all-time classics is stunning so in my opinion that should positively affect the good:mediocre ratio.

but wouldn't Peter, even if given that subjectivity, the ratio N still apply?

Jesse - you write: "the downside of his prolific career is that there were a few mediocre moments" and that's my point. If that's the case, you have a lower N. So be it. The bell curve may destroy the formula itself

I agree with Peter that this is highly subjective; nothing with the words "good" and "mediocre" can ever be objective, in my opinion.

For example, take Steppenwolf, a "one hit wonder" for many people. I absolutely love most of their music, but probably most people will only know "Born to be Wild", so I'll most likely give them a much higher N than the average person will.

I would use a multiplier to weight the total number of good tracks slightly higher than bad ones, so that 10 great / 1 mediocre < 100 great / 15 mediocre, eg, N = G^1.15 / m where (G=great, m=mediocre)

I've actually had this rolling around in my head since you first mentioned it to me, and one question keeps coming back: does the total volume of recorded work need to be factored in to the equation?

Operation Ivy, for example, has an extremely high N -- arguably higher than Zep -- but with only about 1/3 the number of songs.

@Camilo - great points. N does require some large sample size as to number of songs. What I wonder is, can N take some subjectivity out of it however? Again, N doesnt say one artist is better than another, just their ratio.

Also, Stevie Wonder might rank higher than Zep. Maybe.

@whitneymcn (1 why didn't your name autocomplete when I typed it here, since you are in this thread) I agree. See my previous comment.

What's the ultimate metric that we care about here? Influence on other artists?

Woody Allen's "quantity theory" is instructive here. From the recent PBS documentary:

"He says that he's working on the quantity theory, which is that if you just keep knocking them out, one picture after another, just keep making them and making them, some of them won't be that great, but every now and then, one will come out good."

archive.feedblitz.com

@whitney one view of N is that is a law of large numbers theory, so that would presuppose some minimum recorded output. Another idea says screw that, Operation Ivy has a high N with a low output but they are just awesome

@kortina - see Jesse's comment above, he too wants a multipier. I think it destroys the whole fomula

Also: 11 songs on On Avery Island, 11 on In The Aeroplane. Badass N, right there.

@jake - so clearly Woody Allen has a low N. More interesting, is that he embraces that

@andy "One view of N is that..." Why am I not subscribed to the scientific journals that you're reading? ;)

Coming back to your Led Zeppelin example, I can see someone liking only "Stairway to Heaven", which is a bit different from their other songs. So, very low N. (Although if I was ranking it, I'd give it a very high N).

@whitney - NMH raises the question - if you have no mediocre songs, is N useless?

@camilo - proper N requires you can evaluate the whole catalogue; if you only know one song if fails

@aweissman I can see how you think the multiplier destroys it, but if you take it to the limits on either end, I feel it breaks down. Better to create 99 great songs with one mediocre, or 1 great song 0 mediocre. Unless you sum the greatness of each song (so that 1 amazing song w/0 mediocre beats 99 very good / 1 mediocre, which may be possible), I don't think I could go with the original version you proposed.

@kortina so then maybe ny N works as an objective measure, but comparing one artists N to another is subjective??

@aweissman, it's certainly *an* objective measure, but not sure it's definitive. Btw, everyone should check out that woody allen documentary mentioned above.

But what good is such a measure if not for comparison?

BTW, I saw at least some of that Woody Allen documentary, great stuff. I think he's always been very critical of his work, he really seems to hate some of the stuff that the world seems to agree is great. Interesting guy..

I'd skip the multiplier, but I might set a minimum cutoff. The Cars first album had a ridiculously high N imo. But there was no way to maintain that level over time.
And while Zep is clearly above almost all, I'd say the Clash are pretty close to them. I'd also have to figure out how to apply a "Bowie rule". From Space Oddity thru Diamond Dogs, N was off the charts. From Tin Machine thru Reality, not so much

The problem with my generation is we don't appreciate the album-view. But in my opinion, that's what separates the great bands from the good ones. Do they make albums that you can listen to front-to-back?

So if you look at Dark Side of The Moon, for example, the opening track can't be fairly evaluated separate from the context of the rest of the album. And it would be hard to say that it alone is a "great" song.

@graubart Yeah Bowie and The Clash were sort of in my thinking too — seems a shame to have greatness hurt by later not-as-greatness. But maybe that's a separate kind of metric altogether?

@joshm That's a great point, I can think of a bunch of bands whose albums are "atomic" (ex. The Mars Volta)

As for the value of N, first I'll say that Pure Research is undervalued.

From there, I agree with the @kortina/@aweissman postulate that it may not be a comparative measure but a discrete measure for each performer -- a measure of internal consistency, perhaps.

That's also why I resist the "great songs" multiplier: that AC/DC is an extremely low-N band does not make their great songs less great, it is simply an attribute of the band.

Another interesting thread being teased out is that "career" musicians emerging from the '60s and '70s are the low-Ns that seem to make people most uncomfortable, but maybe that's because it quantifies the albums that we avoid thinking about.

I'll come out and say it: Bob Dylan? Overall? Low N.

Bob Dylan and Van Morrison and Bruce Springsteen are interesting examples, hugely prolific, some commercial success but huge fan and critic success. Also a reliance on live performance over recorded material which throws N a huge curveball.

Related: anyone reading this branch should go see, "Searching For Sugar Man"

I am not the hugest fan but it would seems Creedence has merit for consideration based on this formula.

@whitneymcn AC/DC is actually kind of interesting because with Bon Scott you get new perfection with N, after it's hit and miss.

I think Led Zep also had the advantage of long songs durations. It's easier to avoid album fill when a 2-record set like Physical Graffiti only has 15 songs on it.
A much bigger challenge for the Ramones, who I think also have an incredibly high N ratio.
Need to also figure out how to factor in changes in band membership. Until Bob Stinson was kicked out of the Replacements, the band had an incredibly high N, but Don't Tell a Soul and All Shook Down killed that.

@jesse re Bon Scott - wonderful comment

@keane - disagree re the N of Creedance

@whitney - here is why N is interesting - Dylan does indeed have a low N - but maybe that's why he is so awesome? N is not the only measure, just one measure

How do you handle divide by zero errors? For example, Danzig-era Misfits.

I'll echo Josh Miller's suggestion to go see Searching for Sugar Man, saw at last week and was totally blown away by both the story and the music. I've been listening to Rodriguez nonstop since.

@aweissman I like the idea of N as one of many measurements. It also helps make it less competitive and more of a shape/attribute. Bands like Fugazi or Hot Snakes rating super high N but with far less wide audience appeal sort of feels right, where Queen has a pretty brutal N but with giant cultural touchstone significance to the highs.

(That paragraph might sum up my taste more than any other I've ever written...)

@adam - ahh ha. Maybe the denominator needs to be total songs (ie, great songs/total songs)

With Zeppelin and Stevie Wonder you have to take into account the virtuosity factor. Even though e.g. Zep wasn't a jam band, you are constantly aware when you listen to them (studio or live) that, until 75 or 76 at least, they could have plugged in and jammed and been mind-blowingly badass any day of the week, on demand. Which is precisely not dependent on having "great songs" in the way Dylan has "great songs". Even though some of Zeppelin's songwriting is fantastic, it's exactly their ability to transcend the material, on demand, that makes them amazing.

Different from 'quality of live show'. AC/DC plays a great show. Not in the same league 'musically' though.

I do think you need to factor in output. Zep is a good example given the sheer output and high number bona fide hits. Don't forget Coda though. Beatles are likely even higher though. Bands like Neutral Milk, Olivia Tremor Control with two, Jeff Buckley with really one, didn't have a chance to take chances, grow old and make mistakes. It seems like 3 records should be a minimum and then you'd start to see artists like Galaxie 500, Nick Drake begin to stand out.

Perhaps I'm thinking too far ahead, but I'm having trouble understanding the utility of something like this. If you don't incorporate some kind of consensus data, I see it being limited to 2 areas:

1) Individual. You rank songs on a 'greatness' or likability on a media device or within iTunes or some other digital library, which will determine an overall artist N value for music you listen to. How this data is used/implemented is up for discussion. It seems similar to iTunes Genius.

2) Music critics. The reviewer's own N values somehow get incorporated into his/her public profile so that readers would know what kind of slant might be attached to any review that is written.

@andrew to your initial question, who has the highest N -- Derek and the Dominoes. Sort of a cheap answer because they all had more output outside of the one DD album... but the Zep guys all had output after Bonham died, and it would crush their N if you had to factor in all that stuff.

@nillybobles 3) Helpful in starting fights when you're bored.

This is a fun exercise but I'm not sure the simple N value is complex enough. You'd need to build a better algorithm to determine greatness.

Say,

G = greatness
N = good songs/bad songs = quality of an artist's overall catalogue
F = influence

Thus,

G = N * F

One of my most confounding artists that pops to mind is George Harrison. "All Things Must Pass" is canon, but some of his later stuff is utter shit. Yet, "All Things Must Pass" is SO good it out-weighs his clunkers. Also, as a member of another band his F score is off the charts giving him a huge G score despite writing some throw-away records.

@aweissman/@jesse Totally agree: N is just a single attribute, and should not be considered a comprehensive view of the performer or their output. A low-N performer may well have extremely high values in other attributes not yet discovered by Science.

Also, how does this apply to pop artists (i.e. Rihanna) that pay other musicians to write their music? Most artist like this don't carry the same writers from album to album, or even song to song.

@mark, @whitney - G = N * F is very intriguing

@jessevondoom "Here! I have PROOF that they suck!"

@BillyNobles I would submit pop artists almost by definition have a low N

Strangeways, Here We Come is a less than mediocre record . . . otherwise The Smiths hold up nicely in this equation.

Besides formulae, I'm also interested in musical correlations:

1) Number of band members to overall band power (surprisingly inverse correlation)
2) Good drummer to good band (direct correlation)
3) Lead singer that doesn't play instrument to "badassness" of band (inverse correlation with the Jesus Lizard being a notable exception.)

@adams - +1. At some level N is a correlation experiment

@adam how about the presence of one dominant member (or two? beatles, stones, but not many other bands)

@billy I think you need to ignore the producer aspect. I would say Madonna (who falls into this category) has a high N.

@aweissman Ehh its not beneficial to dismiss an entire genre. That practice occurs in other genres as well.

We can't forget that rock and roll didn't even exist at one point, and good/amazing songs and artists were around long before its birth.

Yes and Hip Hop needs to be considered. Nas' Illmatic has off the chart N. Jay-Z has it over his career.

@billy you are right - but N is not a value judgement. I love pop music, but dont think many pop artists have high N

@Andy I think by nature of the term a 'one hit wonder' has a very low N score. but again, this calls to question the relative value of that one amazing single vs. the rest of their crap catalogue. bonus points for charting?

@aweissman I would definitely put Michael Jackson in a category of pop star with huge N and F. Maybe he goes to the top actually with his global reach (though tainted legacy)

@Adam @Whitney agree completely with the N*F. But do you define F as "public recognition" or is it more influence of genres and other musicians? I think of a band like the Replacements - strong N and huge F if you look at influence of other bands. Velvet Underground would be comparable.

@barry btw Velvet Underground have a huge N. Maybe comparable to Zeppelin ;-)

@aweissman I agree completely on Velvets. Though I'd have to exclude Squeeze. Other than Mo Tucker, the band was gone (unless you include Yule).

Thinking of the Velvets, would the N score devalue experimentation? It seems to me that if a band were focused on N, they would stick to a formula once they found it (think DLR days of Van Halen). The Velvets, like the Mats, just didn't really give a cr@p about what anyone said (and weren't selling many albums) so it wouldn't matter. But how do you reward the Beatles for the White Album without penalizing them for why don't we do it in the road, piggies and wild honey pie?

Hmmm.... Is there a way to reward taking risks?

You have to reward one dimension over another. Diversity and risk taking or "hits". Fan popularity or critical acclaim. Each dimension is often counter to the other. I would also throw Talking Heads in the mix with the Velvets.

@aweissman @billy Producers would probably end up having more N than most bands/songwriters. Rick Rubin, Alan Parsons, Quincy Jones, and J Dilla have all influenced entire scenes of artists, creating hit records while maintaining high quality N with few flops. On the flipside, producers like Dr. Luke have massively scaled low quality N for the mainstream with 8/10 songs on each album being mediocre but the hits doing so well it "doesn't matter."

"with 8/10 songs on each album being mediocre but the hits doing so well it "doesn't matter."

Sorta like venture capitalists.

@adam reminds me of a Lefsetz quote: "Ahmet Ertegun was a Venture capitalist. A brilliant risk taker, and he knew how to stomach failure. He also signed artists purely on gut instinct. In many ways these new VC’s and Tech entrepreneurs resemble him."

1. Highly subjective to determine which songs are "good"
2. Strongly biased towards acts with short careers

It might be better to go for periods in an artist's career. Prince, Steve, The Beatles (who'd beat Zep in the other version also), and a few others would rank sky high for certain runs.

@wesley - why the bias if you have short careers? The ratio is not one of longevity per se. And while good is subjective, I hope the ratio takes care of that ;-)

I think there are many variables to consider, one of the most interesting pieces of data to infuse into the equation is how many people "Thumbs Up" or "Thumbs Down" a particular artist in Pandora.

Especially for bands that have been around a lot longer and have the benefit of time on their side this could provide some good tie-breaking data.

This of course begs the question, what is the band that has received the most "Thumbs Up" on Pandora? I'd think that Led Zeppelin could be it, or at least a runner up.

I was scanning through my iTunes ratings earlier to see how many bands were consistently 5s and 4-stars, and how many had variations. I think it showed my "high N" artists - Velvet Underground, David Bowie, Kate Bush (not all will agree I'm sure), Talking Heads, the Clash, Jeff Beck, Ramones, Johnny Thunders. Then there were those which got mostly 3s & 4s with a few 2s and 5s like the Beastie Boys, Bob Dylan, ZZ Top, Depeche Mode and more. Even some who I love, like Iggy Pop, fell into the latter category. Iggy has had some great songs and was a clear innovator, but he's also put out some mediocre music in his career.
It would be interesting to be able to aggregate everyone's iTunes rankings and see the distributions.

@aweissman I agree with @wesley on the short careers. A burst of brilliance is different than someone who consistently delivered quality. Not to dismiss it outright, but there's a lot to be said for delivering a quality second album.

I'll throw a simple ratio into the ring and two adjusting factors:

G = # good songs
B = # bad songs

Ratio = G / (G + B)

Clearly, this scales every band from 0 to 1, with 1 being perfect (no bad songs).

The adjusting factors are found by putting all the songs on a timeline, finding the number of years where 67% of the band's catalog is represented in the fewest years.

Y = # years representing 67%
S = # songs during Y

I picked 67% to represent a loose standard deviation (ignoring outliers), and the fewest number of years to simply be consistent for all bands (I could have used largest number).

These two factors along with the ratio allows comparing the density of creative output and longevity of each band's relevance.

Biased towards short careers because they can have one or two amazing albums and then die, and not have to suffer the rest of their career in your equation.

As a musician myself I know that the more you make songs, the more you have a chance to make a good one, then, my ratio depends on the way I decide to publish all my songs, or only the best ones...

I think that what makes music magical is that not all of it can be captured by a mathematical equation. Sure there is formulaic pop and math rock ;) But what is the equation for Hey Jude? Or the career of Big L? Or Jon Brion's contribution to pop music?

But @wesley, isn't part of the fun trying to capture it? ;-)

Hey, thanks for letting me in, interesting discussion :)

I'm having trouble subscribing to N as an objective measurement. I'm in love with it as a subjective measurement. But the fact of the matter is that my N for The Smiths is just that, it's my N. Your N for The Smiths will vary - which means we're not dealing with an objective truth here.

Now, it's not as romantic a notion but now that most people buy music on a song-by-song basis - couldn't we start to catalogue a variation on N (call it S?) that's based in pure sales data? Sure, it's unfair to new artists, but their figure will form as they gain F and G.

The youth doesn't care about the base of music anymore, they only care about the aspect os the singer. Not get me wrong, I am a melomana.
My real doubt if is there a program to create beats from images?

@tali3sin I think that using sales data will kill the whole point here. Think of the famous Eno quote about the Velvet Underground - they only sold 10,000 albums but every one wh bought one formed a band. Huge, huge N, Low, low S.

While I think Adele has a strong N, I don't think Katy Perry nor Pitbull do, yet they would have a strong S.

Yes, N would be very subjective. But to simplify it, you could start by analyzing your own music collection. Some would have high N, others low. And if you and I compared music collections, it's possible that we would not have much in common. (cont'd)

Cont'd from above. For example, my wife was a disco girl in the late 70s, while I was a punk rocker. So the venn diagram of our joint musical interests is limited. BUT, if we look at those bands we both like, some like Parliament/Funkadelic would be fairly high N, some, like Prince, moderate N (IMO) while others, like the Rolling Stones, lower N.

Now, assume we can look at everyone's N scores, whether using iTunes stars or Pandora thumbs up. I'd be that within fans of a given band, there would be commonality to their N factor.

So, even tho you & I might like different music, music listeners overall should come up with accurate N scores.

In theory...

@Barry Then, how can you tell if a song if good or bad? I think the best way to do this and still being objective is by using sales data.

But I totally agree. The music industry is full of crap and you will be limited by their (manipulated) numbers.

@wesleyverhoeve I agree that attempts to use technology to "determine a hit" will fail. There is no equation for Hey Jude (though perhaps for Revolution #9).
But that's not the goal here. The goal is to use math to assess whether certain performers tend to have a higher ratio of great music to overall music, whatever that music may be.

Looked at it another way - there are certain albums we've all bought where you only listen to the same 2 songs and never want to hear the rest. Conversely, there are albums that you always listen to start to finish. That's something that can be scored and measured. It's not the magic of the music we're scoring, but simply whether perceived quality is consistently high or not.

@MrJJSanti The data that I want probably does not yet exist in a machine-readable form. But it might. I go back to the iTunes stars. While I think most people's iTunes libraries are like Lake Wobegon (most songs get 3-5 stars, few get 1-2), if we could compile everyone's star ratings, then normalize them along whatever curve they use, and aggregate them, we could probably score the N's. An album that only had 500 listeners, but where those 500 consistently rated all songs on that album as 4s and 5s (and where those same listeners rated other songs as 2s & 3s), would have a stronger N score than an album with 50,000 listeners who rated one song on that album a 5, with a mix of 1s, 2s, 3s & 4s for the rest.

Sales data is marketing driven

@graubart You make some really solid points Barry, and I feel like your aggregated iTunes stars system would give us something that is always approaching objectivity - but would never get there. I'm cool with that. Part of me feels that this is what Pandora and Last.fm have been doing since they were born, and it's part of the reason I love them. They're just not assigning bands a number, rather spending their time helping us find new tunes.

Speaking of Last.fm what if we ditch sales data, and used listens? (excepting radio, I suppose) Sales data *is* marketing driven and in the age of Rdio, Spotify, Pandora, iTunes Match, Grooveshark, etc, sales data is increasingly less relevant.

Not sure it makes it any less subjective though. Hmm.

To be honest, regardless of whether it's objective or subjective, N is a really great idea. As if Metacritic scoring worked on every track, took account of every review and every rating.

@tali3si I think the idea of using last.fm listens would be terrific.

I just scanned my listens for the past 12 months: last.fm
and it's easy to see some bands where I'm listening to the whole album and others where it's just a few tracks.

Of course, my individual stuff is skewed by the playlists I've created, but that reflects the reality of how I listen. And if we aggregate, that all should get normalized somewhat.

OK - who wants to use the Last.fm API to hack this together?
last.fm

I think only using Last.fm for this is a poor idea, but it's certainly a start. This could probably easily be pieced together in python (given the number of results for last.fm on PyPI: pypi.python.org). If anyone has any experience with pandas (pandas.pydata.org) this would be very easy.

On the other hand, someone much more familiar with R could easily figure out the proper API calls and aggregate the data into tables which would be far more easily manipulated.

We could use this to algorithmically determine what is common among these songs that makes them popular: time signature, key, variation of "riffs" (for lack of a better word), &c. Using this we could apply it to other artists and see how well it holds.

I would second Zep, Neutral Milk Hotel and Michael Jackson. I would add New Order, if you're into that sort of thing. Even their mediocre, late period album cuts are still kind of amazing, genius pop songs.

As long as we're continuing to add bands to the list of bands we believe to have a high value of N, I'd suggest Radiohead and Sigur Ros. I'm also a Zeppelin fan so you'll hear no argument from me about them. And I believe I saw Pink Floyd mentioned, but if not, tack them onto the list on my behalf as well.

Last.fm isn't widespread enough to do the idea justice, but as you say, it's a start. It's a pity I'm not much of a programmer. Might see if I can wrangle a friend into taking a look at this.

@paul that would be awesome.

I might have sone time to work on it as well but so far a problem might be that the python wrapper around the API is outdated. I cannot test the idea without a last.fm account it seems. If this is taken seriously though and open sourced (under whichever license) I would be glad to assist.

Unfortunately my friend has too many projects going on at the moment. How terribly selfish of them ;)

Ian, we can just start up a Last.fm account - or if one is required, I'm happy to sacrifice my long term account to the cause. Almost 32,000 listens attributed just to me.

As @tali3sin says, easy to start an account. Or, if we need an "active" one, we can use his or mine. Worth a quick test.

Having the global N and the personal N for bands is probably quite possible, if the API lets you access "all listen data for a band" and "all listen data from a user for a band"

Sorry, kind of busy myself. I'll see what I can do, but I was under the impression that last.fm was a pay-for service.