| For context: I got inspired by this forum thread: https://myanimelist.net/forum/?topicid=2169128 and decided to take it to the next level. Every anime on MAL has a favorite count, showing how many people marked it as their favorite. Favorites Ratio tries to tell you the percentage of people who who watched an anime and marked it as their favorite. We use watching and completed statistics to calculate this metric, as it is very likely most people who favorited an anime are either watching it or have it set as completed. I ran calculations on all anime that were available on MAL at the time of running this, 28922 entires. Down below you have the Top 100 using a weighted average, along with two additional Top 10s i decided to throw in for fun (ONA & Hentai excluded). Formula: R = (favorites / v) * 100 v = watching + completed m = 5000 C = The site-wide average favorites ratio of all anime. Weighted Ratio = (v / (v + m)) * R + (m / (v + m)) * C Not perfect (and can never be), there are many different ways you can approach this with varying outcomes, like tweaking the parameters to allow mainstream anime to take up more or less space. (never used tables here before, so praying this is going to look OK) If there is something you'd like to see that i didn't include here, let me know. I'm not going to publish the whole list. What do you think? Is this an interesting way of ranking anime? |
| I'm not much of a data and statistics guy but the information is appreciated and very useful. |
| Appreciated list! Although I'd say m is a little bit too high perhaps, it will punish lesser known titles quite a bit. For the uninitiated, this would be the statistics equivalent of adding 5000x site-wide average favorites ratio to all shows. An easier analogy. if it was scores, it'd be adding 5000x global average score to every show. Let's say site average is 6. So a show with 5000 members who all rated it 10 would become (10+6)/2=8, which probably isn't the outcome you'd like, because by 5000 members the uncertainty is already fairly reduced, reduced enough to not warrant a whole 2 point reduction. If we assume the uncertainty curve takes a 1/√n shape, the highest m that I can get behind is like 500. For reference MAL's m is 100. |
 |
| This kind of data is useful because it's not a coincidence that some anime are ranked much higher by favorites ratio than by score, just as two anime with similar scores can have a vastly different percentage of recommended reviews. Numbers must always be interpreted. |
| It looks great but- I'm rubbing my eyes, tell me I didn't miss it in there somewhere- but the GREATESTEST ANIME EVAR MADE, the GOAT Frieren is not even in the top 100??? Hahahahahahahahahahahaaaaaaaa!!! That's so fucking hilarious It just proves even more how utterly fake the score on that anime is. All the morons who created fake accounts to push it to #1 didn't bother faving it apparently- but if they've read this they're out there furiously clicking now aren't they? |
| Nice seeing LotGH still going strong. @SuperAdventure It's 15th. |
| Neat ranking, I actually did something similar a while ago: I ranked the companies by favorites per number of anime. #1 is Studio Ghibli, followed by Studio Bind. Minor nitpick: The ratio is actually a percentage. |
ZarutakuSep 11, 4:21 PM
No, this isn't my signature. |
| @Nirinbo Favorites is an interesting metric, i agree, but i wouldn't trust the numbers too much, I've seen a fair bit of odd data combing through the raw numbers. I won't go any deeper into it, just putting it out there. @Zarutaku Yes, you're right @Mantighast @Auron No specific reason, but 5000 is indeed harsh. We can lower that, lets try 100. |
| Very interesting to see. Thanks for putting this table together. |
  |
Reply to SomeNewGuy
@Nirinbo
Favorites is an interesting metric, i agree, but i wouldn't trust the numbers too much, I've seen a fair bit of odd data combing through the raw numbers. I won't go any deeper into it, just putting it out there.
@Zarutaku
Yes, you're right
@Mantighast @Auron
No specific reason, but 5000 is indeed harsh. We can lower that, lets try 100.
Favorites is an interesting metric, i agree, but i wouldn't trust the numbers too much, I've seen a fair bit of odd data combing through the raw numbers. I won't go any deeper into it, just putting it out there.
@Zarutaku
Yes, you're right
@Mantighast @Auron
No specific reason, but 5000 is indeed harsh. We can lower that, lets try 100.
| The top 100 with m=100 actually seem more interesting, because the hidden gems are ranked higher than before. |
No, this isn't my signature. |
Reply to SomeNewGuy
@Nirinbo
Favorites is an interesting metric, i agree, but i wouldn't trust the numbers too much, I've seen a fair bit of odd data combing through the raw numbers. I won't go any deeper into it, just putting it out there.
@Zarutaku
Yes, you're right
@Mantighast @Auron
No specific reason, but 5000 is indeed harsh. We can lower that, lets try 100.
Favorites is an interesting metric, i agree, but i wouldn't trust the numbers too much, I've seen a fair bit of odd data combing through the raw numbers. I won't go any deeper into it, just putting it out there.
@Zarutaku
Yes, you're right
@Mantighast @Auron
No specific reason, but 5000 is indeed harsh. We can lower that, lets try 100.
SomeNewGuy said: Favorites is an interesting metric, i agree, but i wouldn't trust the numbers too much, I've seen a fair bit of odd data combing through the raw numbers. I won't go any deeper into it, just putting it out there. Yeah, I said that numbers need to be interpreted because there are some outliers that should be ruled out (the Jujutsu Kaisen PVs obviously aren't among the best anime of all time). I'm not even talking about one person's taste: in terms of general reception anime like Nana, The Tatami Galaxy, Utena and Lain have always been considered better than many anime rated higher than them. The % of favorites is showing something that scores alone couldn't (I mean raw scores, maybe "recalculate scores after only taking into account votes given by users with more than X completed anime and a mean score below Y" could give us similar results, who knows?). |
Reply to Shizuna
| @Shizuna Ah I see now I was looking for just Frieren and not Sosou no Frieren, thanks. My eyeballs hate reading spreadsheets In any case, I wld argue that favorites are a more accurate way to rank on the site. It's a 15th most favorited anime, great. 15th is a long way from # 1 |
Reply to SuperAdventure
@Shizuna Ah I see now I was looking for just Frieren and not Sosou no Frieren, thanks. My eyeballs hate reading spreadsheets
In any case, I wld argue that favorites are a more accurate way to rank on the site. It's a 15th most favorited anime, great. 15th is a long way from # 1
In any case, I wld argue that favorites are a more accurate way to rank on the site. It's a 15th most favorited anime, great. 15th is a long way from # 1
SuperAdventure said: I wld argue that favorites are a more accurate way to rank Maybe The powerful thing about favorites is that there is only a finite number of favorites a user can give out, unlike with mean score where you have "unlimited" 10/10s, it is not unusual to see users with 20 or 30+, 10/10s. It is important to keep in mind though that different people might have different definitions of what a "favorite" anime is. You could give an anime a 8/10 but still want to have it as your favorite over a 10/10 anime in your list. So a favorite can't really be treated the same way as a 10/10, but it is a positive signal nevertheless that there is something special about the particular anime that the user decided to favorite, it is something worth paying attention to. I think the correct answer here might be to create a new metric where favorites and mean score can be used in symphony to rank anime. I experimented a little with this, give a 60% influence to mean scores and a 40% influence to favorites, lets call this new metric a "resonance score". Then in that case, a Top 100 would probably look something like this: |
SomeNewGuySep 12, 8:02 AM
Reply to SomeNewGuy
SuperAdventure said:
I wld argue that favorites are a more accurate way to rank
I wld argue that favorites are a more accurate way to rank
Maybe
The powerful thing about favorites is that there is only a finite number of favorites a user can give out, unlike with mean score where you have "unlimited" 10/10s, it is not unusual to see users with 20 or 30+, 10/10s.
It is important to keep in mind though that different people might have different definitions of what a "favorite" anime is. You could give an anime a 8/10 but still want to have it as your favorite over a 10/10 anime in your list. So a favorite can't really be treated the same way as a 10/10, but it is a positive signal nevertheless that there is something special about the particular anime that the user decided to favorite, it is something worth paying attention to.
I think the correct answer here might be to create a new metric where favorites and mean score can be used in symphony to rank anime.
I experimented a little with this, give a 60% influence to mean scores and a 40% influence to favorites, lets call this new metric a "resonance score". Then in that case, a Top 100 would probably look something like this:
SomeNewGuy said: Maybe The powerful thing about favorites is that there is only a finite number of favorites a user can give out, unlike with mean score where you have "unlimited" 10/10s, it is not unusual to see users with 20 or 30+, 10/10s. It is important to keep in mind though that different people might have different definitions of what a "favorite" anime is. ^This is exactly why it would be a better ranking system- due to the limit they would not be able to warp the rankings by just scoring 1000s of things '10' they never even watched. Each account would have a limit of 10 anime (or 20 for premium) they could do that. It would limit abuse, which is exactly what happened with the Frieren thing. "It is important keep in mind ...different people might have different definitions..." This subjectivism is exactly the same with scoring. Each user has different definitions of what a 10 (and 1-9) is. So it would be no difference. |
| Tatami Galaxy being in top 10 is pretty interesting. Don't think it can just be attributed to a small number of hardcore fans like Utena or Ashita no Joe 2. Also interesting that Detective Conan is so high up. I wonder if lots of people moved it out of their Watching list as they moved on, but kept it in their favorites list. Jujutsu Kaisen PVs being at #25 is amusing. Wonder what happened there. The PV looks pretty ordinary. |
| Statistics never work for me when I an searching for entertainment. |
More topics from this board
» A Cartoon you wish was Anime instead ?tchitchouan - 11 hours ago |
37 |
by Captain-577
»»
14 minutes ago |
|
» My dress up darling no romanceNilyaaa - 27 minutes ago |
0 |
by Nilyaaa
»»
27 minutes ago |
|
» Were battles much longer in older anime?fallout45 - 28 minutes ago |
0 |
by fallout45
»»
28 minutes ago |
|
» "Anime questionnaire" ( 1 2 3 4 5 )llene - May 31, 2021 |
234 |
by otakupro68
»»
30 minutes ago |
|
» Kinks you find attractive in anime, but would freak you out in real life?NotYourLittleSis - Today |
18 |
by Captain-577
»»
37 minutes ago |