A Novel Approach to Determining the Most Successful Songs of All-Time on the Gaon/Circle Chart

It's definitely a good try. We'll never find a way to do it in a completely fair way but as you said this kind of analyzes usually point out to the same songs so they're definitely among the biggest.

I wanted to ask though, you're using just the data of the top 100 charts right? I was a bit surprised to see Cherry blossom ending at #12 (it would be #1 in my personal ranking), but it's true its success is mainly focused on the recharting on springs, so it doesn't make it into the top 100.

It got 8M downloads confirmed by Circle though, so I'm not sure if you're counting these things. There're more like TTN with 7M thanks in a big part to a crazy longevity in the top 100-200. In general there're several of these cases and I think that's the reason why maybe the 2010-2015 are fewer in general, the charts were more active back then so songs used to leave the top 100 faster.

This is just me being a bit picky though lmao, the analysis is pretty good anyway, congrats!

Also surprised to see LD and Hype boy in those positions while still charting pretty high in the charts, ig they're just way bigger than the other songs released in the last couple of years.

i ilke how ive and nj are always side by side. it's cute

It's definitely a new way to look at songs' success across eras and it is fairer than just comparing downloads/streams/points but it still seems to favour songs released in the 2020s and the latter half of the 2010s.

Of course BTS's 4 sotys appearing are no surprise

We need to assemble a K-Pop analytics team so that a fool-proof, almost universally accepted formula for song comparison can be created. This was a great solo effort at least, and it's more than what most are ever willing to do. I know I'm not mathematically competent enough to even attempt it.

But I wish we could find a fair and objective calculation method and metric, so these usual lazy, agenda driven narratives that happen in song comparison discussions can be put to rest for good.

I'd almost be willing to pay a group of stat nerds to do it.

Akmu 2nd I know that's right!

On closer reflection, a group that is noticeably absent from the list is BLACKPINK. I said I don't have the qualifications to critique your calculation method, but I would question the math behind the end result if BLACKPINK's biggest hit song in Korea by a mile, and one of the biggest girl group songs ever, D4, doesn't make a top 50 ranking.

Quote from YamaChwan

We need to assemble a K-Pop analytics team so that a fool-proof, almost universally accepted formula for song comparison can be created. This was a great solo effort at least, and it's more than what most are ever willing to do. I know I'm not mathematically competent enough to even attempt it.

But I wish we could find a fair and objective calculation method and metric, so these usual lazy, agenda driven narratives that happen in song comparison discussions can be put to rest for good.

I'd almost be willing to pay a group of stat nerds to do it.

There isn't a way to make a completely objective list though. It's not just about kpop but the music industry in general is in constant change, and those changes don't have any correlation between them. You can't compare the physical sales of the 90s with the downloads we started having later, now the streaming era, and how would you rate the popularity of songs in Tiktok nowadays?

That's why I'm more a fan of tiers to do this kind of things, you're not ever going to know what song is IU's biggest song, Good day? Friday? TTN? Any of them have some reasons that could make them the biggest, however it's obvious they all would belong to a "top tier"

Quote from YamaChwan

We need to assemble a K-Pop analytics team so that a fool-proof, almost universally accepted formula for song comparison can be created.

Yeah, that's never going to happen in this fandom. 🤣 Also, I think it's pretty much waste of time, because most fans don't care about these stats, unless their faves are in it. Most of them are ignoring pure facts too, not just lazy, so there's no point doing it.

Quote from uwuwu-98

There isn't a way to make a completely objective list though. It's not just about kpop but the music industry in general is in constant change, and those changes don't have any correlation between them. You can't compare the physical sales of the 90s with the downloads we started having later, now the streaming era, and how would you rate the popularity of songs in Tiktok nowadays?

That's why I'm more a fan of tiers to do this kind of things, you're not ever going to know what song is IU's biggest song, Good day? Friday? TTN? Any of them have some reasons that could make them the biggest, however it's obvious they all would belong to a "top tier"

That's a valid perspective, and one I'm aware of, but I also don't believe the calculation method itself wouldn't also need to undergo iterations and evolve over time. The closest parallel I can think of is sports.

Around the early 2010s, an advanced analytics surge happened, and professional NBA teams were hiring less scouts and talent evaluators and more statisticians and data analysts. In 2024, every professional team has an analytics staff that finds ways to quantify and compare players, and we have metrics like Player Efficiency Rating, Real-Adjusted Box Plus Minus, Win Shares, Value Over Replacement and the list goes on.

And basketball is a sport that has changed massively in 10 years.

I just want to manifest an equivalent to that in K-Pop. I'll even consider funding it! Because I'm very curious what a group of really smart people could come up with.

I don't mean to hijack your thread, OP. The kind of work you put into this data is what I'd like to see more of, regardless of how I feel about the end result.

LYW being that high is making me question the methodology

Saving that download to streaming difference chart I keep telling people that Growl was released EXACTLY at the crux of the streaming era and it’s position on the digital chart isn’t truly representative of how actually big the song was.

It was #2 on the streaming chart something that’s absolutely incredible while it was #56 on the digital chart. The Gaon/circle measuring methodology has gone through so many changes since then lol.

Anyway I don’t really think anyone can find a full proof way to rank songs imo the year end chart isn’t that great a chart. But good way to gain some insights.

I didn't quite understand the method but there are no songs from 2010-2013 in that ranking. Wouldn't that mean that the hits from those years can't absolutely not compete to songs post 2014 songs (which would be weird), or the method didn't quite manage to balance it out

interesting

this method is sorta what I was talking about since it ranks the song compared to one's peers

it doesn't matter the raw numbers it's how one does compared to one's peers that matter

now based on what you've shown it seems that the data for 2023 is either not out or no group achieved anything significant with any songs in 2023?

Always impressed by OP. Thanks and it was interesting to read.

How did ddu du ddu du not make it? It's still the biggest gg song in cumulative on circle. Tho it'll be overtaken

Quote from David33

Yeah, that's never going to happen in this fandom. 🤣 Also, I think it's pretty much waste of time, because most fans don't care about these stats, unless their faves are in it. Most of them are ignoring pure facts too, not just lazy, so there's no point doing it.

I do not see my faves on the list.

This method is Garbage!

Also why no as if it's your last?

Quote from wsoet

Founded in 2010 by the Korea Music Content Association, the G‍aon Chart (now known as the C‍ircle Music Chart) is the premier source for tracking the digital performance of songs in South Korea. I've always been curious as to the most successful songs of all-time on the chart, but making this determination has always been difficult for a couple of reasons.

....

i've been thinking about doing exactly this, but for melon uls of which we have reliable data since at least 2019 on guysome.

There are a couple faults i see with this method tho, and i dont see a easy way to take account for it...

In 2013 the first price hike came for downloads then it started to decline until it basically became void and mostly used by fandoms to help charting around 2018-2019.

With the way you combine streams and downloads pre 2018 here it falls apart a little, because a download for gaon is worth much more for their scoring system than a stream is and the exact algorithm isnt known (their logic is the score should reflect the revenue of a song, which doesnt really make much sense when they use filtered streams aka uls and not actual unfiltered streams but, alas).

They have also changed their algorithm for it at least once since they reimplemented it in 2018. They actually had a digital chart with digital point system years before 2018 as well, but it didnt work the same and none used it when streams and ul numbers were still directly available

Another factor for digital points, that the normalization doesnt really take into account is that songs on average today stay much longer in top 100 then they did years ago and is bascially gone up year by year since 2019. This means news songs (especially if only looking at top 100) will acrue much more digital points over time.

Looking at the top songs the vast majority are 2019 or later, and it has a couple of big omissions especially from 2015 and earlier when downloads ruled.

I think there is no easy solution to combine pre 2018 data with post 2018 data. As such think it would be interesting to see one for all songs post 2018 without combining it with data before that that year into it, and maybe do some similiar normalization of average chart time to take that into account as well.

Similar same could be done for downloads and streams pre 2018 (but probably best to keep them apart)

But regardless it all makes for an intersting data point and great job!

When I read "Friday", I thought it was gonna say "Rebecca Black".

Quote from uwuwu-98

There isn't a way to make a completely objective list though. It's not just about kpop but the music industry in general is in constant change, and those changes don't have any correlation between them. You can't compare the physical sales of the 90s with the downloads we started having later, now the streaming era, and how would you rate the popularity of songs in Tiktok nowadays?

That's why I'm more a fan of tiers to do this kind of things, you're not ever going to know what song is IU's biggest song, Good day? Friday? TTN? Any of them have some reasons that could make them the biggest, however it's obvious they all would belong to a "top tier"

I won't say about others, but the only song of IU worth considering is Bbibbi which will probably remain as her only song which exceeded 250m views.

Youtube views , more than Gaon which no one really cares outside of Korea, is a good empirical standard

Quote from WhyKnock

Youtube views , more than Gaon which no one really cares outside of Korea, is a good empirical standard

Only if you're excluding the paid views!

Quote from wsoet

This phenomenon is something that stood out to me when I was putting together this analysis, and so I'm glad to see that you all noticed it as well. To look further into this issue, I calculated the standard deviation for each year's set of normalised scores, and the results are shown in the chart below:

STDEV_Normalised.png

As we can see, standard deviations have been steadily rising, which means that there's been more and more variability between normalised scores, even if the *average* normalised score remains at 100% for each year (which it will, by design, due to my methodology). Or, to put it more simply: the best performing songs have been able to outperform their peers by larger margins as the years have progressed.

I'm not sure how to correct for this. It seems like I would have to "standardise" the standard deviations somehow, so that they remain roughly constant across all years.

Display More

Apart from the average length of charting, i think a big part of the problem it is this calculation:
"For example, if a song had 1.5 million downloads in 2010, and the average song in the t‍op 100 had 1 million downloads, I would assign a value of 150% (= 1.5 million ÷ 1 million) to that song."

Neither streams nor downloads can be directly translated to gaons 2018- digital point system without knowing their exact formula as its not a 1:1 factor and one download is worth alot more points than a stream in their calculations.

For example (made up numbers) say one song have 10 million streams and 100k downloads. This would translate to 10,1 digital points in your calculation for songs pre 2018. But those 10 million streams would be worth lets say 50m gaon points while 100k downloads alone would be worth 100m gaon points.

This means that by your caclulation, if another song have 10million streams but 0 downloads, the score would only have a difference of 0,1m points between them, while in gaon points, it would have a 100m difference.

Downloads are the single reason why some artists (kang daniel, lyw etc) have been able to rank #1 on gaon in their first week despite never even having entered top 20 on melon for example. The fandoms simply mass buy (download) the songs on release week which inflates the digital points way above whats normal versus their UL numbers.

The reason why more recent songs chart better are also really influenced by the change of the market regarding downloads and streams. When we start the decade in 2010 downloads were the main way to consume music, you download the song once, now downloads are completely dead and the market is 99% streaming unless you've a big fandom support.

Because of this songs before lacked any longevity, let's say a huge hit back then released in June could reach #1 in the yearly chart but the charts moved super fast because downloads were more dominant, so it was impossible for that song to even make it into the next year top 100 unless it became a riser or a seasonal song. Now a huge hit released in June might have it a bit harder to reach #1, maybe it can just make it into the top 5-10, but you'll see it the following year in the top 30 or top 40, and it can also make it into a top 100 of a 3rd year perfectly, accumulating points from several years that older songs have basically locked for the system of music consumption itself.

I don't think this has solution tbh, downloads are always going to be an issue if we want to compare songs released far in time, even after the price change in 2013 they were still really relevant until 2019 and like Kreatin said they can't be translated directly to digital points.

Appreciate the effort but this list still feels way too heavily leaning towards post 2019 songs... Just looking at IU's entry songs in this speaks A LOT. One cannot convince me that celebrity lilac hold my hands or even eight (and most of the songs on this list for what matters) are bigger hits than Good day, you & I and friday.

It's impossible to compare like this imo cause this doesn't take into consideration how streaming, or whatever happened, influenced longevity and made charts stall

i mean i like IVE and everything but for after like to be above cheer up or every bp song looks pretty wrong. back to the drawing board i say

Quote from Cold_Wave

i mean i like IVE and everything but for after like to be above cheer up or every bp song looks pretty wrong. back to the drawing board i say

I mean op didn’t calculate Al differently.

After like has better 27 weeks on melon top 10 and was the in the top of 2023’s biggest hits and is still charting

Blackpink biggest hits are DDDD but most of its digital points are from 2018, AIYLT despite being their longest charting song haven’t charted in the top 100 for years.

Like no one is denying cheer up is bigger but songs back then lacked longevity.

Quote from wsoet

I just checked my numbers again, and here's what I have for Lim Young-woong's two songs on the list:

사랑은 늘 도망가 (Love Always Runs Away)

#86 in 2021: 234,640,416 points ÷ 368,719,639 average points = 63.64%

#05 in 2022: 635,225,646 points ÷ 347,216,653 average points = 182.95%

#22 in 2023: 374,798,361 points ÷ 293,478,667 average points = 127.71%

Total Normalised Score: 374.29%

이제 나만 믿어요 (Trust in Me)

#50 in 2020: 391,308,836 points ÷ 426,902,573 average points = 91.66%

#35 in 2021: 391,197,331 points ÷ 368,719,639 average points = 106.10%

#51 in 2022: 312,888,152 points ÷ 347,216,653 average points = 90.11%

#75 in 2023: 213,671,641 points ÷ 293,478,667 average points = 72.81%

Total Normalised Score: 360.68%

Given these numbers, it seems like it's a just a case of his fandom's dedication allowing the songs to have impressive longevity.

Display More

The increased longevity of songs is one thing that is very difficult to amend for as it’s down to various factors. Some as pervasive as the fragmentation of music consumption all the way down to Melon changing their charts here and there. Timelines all over the place.

Quote from wsoet

I just checked my numbers again, and here's what I have for Ddu-Du Ddu-Du:

#5 in 2018: 940,214,913 points ÷ 545,784,501 average points = 172.27%

#92 in 2019: 312,443,592 points ÷ 517,358,740 average points = 60.39%

Total Normalised Score: 232.66%, which results in a rank of 69th (nice)

The reason why the song didn't finish with a higher ranking is likely because its placement in Gaon top 100 dropped significantly in 2019. Keep in mind that many of the highest-ranked songs have stayed in the top 100 for three or more years.

Display More

Anything on blackpink as if it's your last? Even red flavour, dance the night away? Or they don't rank in your rankings?