What is the best predictor of an upcoming anime’s quality?

This post was written by Dark_Sage. He is Dark_Sage.


As the new season nears ever closer, I think it’s worth investigating what the best indicators of an anime’s quality might be. (My current system has not been working too well.) Unfortunately, the only mathy way I could think of doing this is with MAL’s scores, so prepare yourself for some ratings-related retardation.

Before We Begin

Safety first!
Safety first!

What you’re gonna be looking at on these spreadsheets is how the MyAnimeList rating compares to the average score (from the past two years) of a given anime’s director, series compiler, and production company.

Historical scores were pulled from the past two years (2011-2012). I figure it’s more accurate to use what someone’s done recently as opposed to something they did in 1995.

Ratings/adaptation info pulled from MyAnimeList, series information pulled from AnimeNewsNetwork. Data pulled December 16th, 2012 and December 17th, 2012. (That’s some fucking ace-tier citing right there.)

Oh, and I’d like to mention that although I’m presenting this as some super-scientific mathalicious stuff, it’s 100% statistically unsound.

Google Docs if you wanna fuck around with it: https://docs.google.com/spreadsheet/ccc?key=0AgDvnCqWkRhmdExVa01nQXFZWi1fOGJSZi1PRWg1TkE


“Key” Animation Staff/Companies:

Director – Some guy.

Series Composition – Script overseer.

Animation Company – Company with which to of the animation.



Wait, What’s Happening Here?


tl;dr: Stats stuff to find a decent way to predict an anime series’ quality before it airs.




Fall 2012

Fall 2012 Anime Stats Part 1-1

Fall 2012 Anime Stats Part 2-2


Fall 2012 conclusions:

It appears that series composition is the most accurate predictor of series quality, with animation companies being the least accurate predictors.

When data is bunched together, “director+series composition”, “series composition+animation company” and “director+series composition+animation company” are all equally the most accurate predictors of quality, even more accurate than just looking at series composition by itself. “director+animation company” is noticeably the least accurate.

From here I guess we can just say that series composition is the primary factor in a given series’ quality. But hey, why not bring another season’s data into the mix just so we can be sure?


Summer 2012

Summer 2012 Anime Stats Part 1

Summer 2012 Anime Stats Part 2


Summer 2012 conclusions:

It appears that animation companies are the most accurate predictors of series quality, with series composition being the least accurate predictor.

When data is bunched together, “director+animation company” is the most accurate predictor of quality, somewhat closely followed by “series composition+animation company” and “director+series composition+animation company”. “director+series composition” is the least accurate.

For Summer 2012 it looks like animation company is the primary factor in a given series’ quality.

Well, fuck.

That isn’t quite what I said for Fall 2012. In fact, it’s almost the complete opposite. What if we bunch the seasons together then?



Summer 2012 + Fall 2012

Summer and Fall 2012 Anime Stats

When both seasons have their data smooshed together, a different kind of pattern emerges.

While “animation company” by itself is a slightly better indicator of quality than “director” or “series composition”, when you combine series composition and animation company, you get the most accurate predictor. Combining director along with series composition and animation company gets you similar results, but the range of scores is higher, indicating that just sticking with series composition and animation company is a more stable predictor.

If your season predictions are usually as fucked as mine are, taking a look at those two aspects of upcoming releases could likely help you narrow down what you’ll enjoy watching the most. And yes, I’ll be relying on these conclusions for my Winter 2013 preview post.



Original Anime vs. Adaptations (Manga/Light Novel/Visual Novels/Anne Frank’s Diary)


Thought I’d see if there was a major difference between adaptation sources in terms of quality and what you could expect from upcoming series based on it. But it looks like there really isn’t a difference, at least between manga and light novel adaptations. What did catch my eye, though, was the “original” score.

Seems like anime companies do all right when they have something to copy from. But warning bells should go off when the animators talk coming up with something on their own. Leave it to the writers, animators.

Oh, and that 7.38 for Anne Frank’s diary? Guilt ratings; the movie sucked worse than her hide-and-go-seek skills.



Final Conclusion


MAL’s userbase has shit taste.

32 thoughts on “What is the best predictor of an upcoming anime’s quality?”

    • “And yes, I’ll be relying on these conclusions for my Winter 2013 preview post.”

      I’d say this is more of a prequel than a part 1, though.

      • What I like most about your previews is how you read bits of the source material if any exist. Most other previews rely on plot synopsis and staff listings exclusively. It’s nice to have this third element.

        Doing all that extra reading (of source material) obviously takes time. Any chance you’ll get the finished product out this week? (I do not mean to pressure you…the first show doesn’t even begin airing for another 15 days)

        • You’ll have it this week. I get the feeling I should start looking at some oldsubs before I start on my next major article though.

    • I don’t think the 6.5 to 8.5 is really the main issue here. Rather, it’s that I didn’t have enough data to use or enough factors of comparison (well really, those are just a few of the issues though). If I were querying a database here, I guarantee I could get numbers that would say something of value. But with how much I had to leave out, it’s more of a conversation starter than a revolutionary new way of looking at things.

  1. I generally just look at the anidb pic to decide if a series will be good or not. Surprisingly, I’m right a lot of the time :o

    Surely PVs are a better way to see if an upcoming series is going to be any good? They’d be more accurate than doing statistics via MAL >.>

    • PVs don’t usually help me decide a damn thing, unfortunately. I include them in my preview posts though because they’re an interesting and easy way of breaking up the monotony of text.

      Also, I didn’t mean to imply that MAL’s stats would be driving my personal opinions. That’s stupid. What I meant was that applying my prior experiences to the past works the staff/companies worked on may be more accurate than my previous process of “oh, that paragraph sounds interesting; to the top of my list!”

  2. Studying for the next preview, eh? Don’t worry, I don’t think many people will be reading for your ever accurate predictions anyway.

  3. I use 4 factors to help me decide:

    a) Promotional art. If the main purpose of the promotional art is to show teenage girls showing lingerie, the show gets dropped instantly. Fanservice is an added bonus, not the purpose of a show.

    b) Synopsis. Of course, the plot has to interest me if I want to watch a show. Lack of a plot means instant drop. This makes me drop almost every comedy-oriented show, which spares me from cringing at untranslatable puns and what dumbass subbers do with them.

    c) Previous knowledge of the source material. If I know the manga/LN/VN/comic/novel/whatever the fuck is good, then the show goes on my list. If I know it’s garbage, then it’s a drop. (This factor has betrayed me sometimes, though. For example: Zetman.)

    d) Checking if BONES and/or Square Enix aren’t involved. I hate BONES endings, Xam’d was tripe, and I’m still bitter at Squeenix for fagging up FMA a la Final Fantasy. I mean, I was young and impressionable. And I was betrayed.

    Usage of these 4 factors makes sure I have an 80% chance of getting good shows, according to my personal preferences. However, they’re also a double-edged sword: I end up choosing at most 4 or 5 shows each season, and there are times when the whole season gets dropped.

    Maybe I just don’t like anime, but I haven’t yet realized it.

  4. I use the simple process of looking at the cover art and the synopsis to decide whether something will be interesting.
    Then I forget all about it till the new season starts and just watch the first five minutes of every show to make up my mind.

  5. >Company with which to of the animation.
    >even more accurate than just look at
    >Summer 2012 is looks

    Pretty long post just to say MAL’s userbase has shit taste, though. Seems like a waste of effort.

  6. I really believed you were going for something here, I really did – then you started with Fall + Summer averages and ranges and that >= series composition – director + Oreimo and my brain closed.

    National J-Animedesu Security should hire you, though!

  7. I’m usually waiting for the first couple of reviews here and then use the screenshots as main indicator of whether I might like the series or not.

    Don’t think this will work for you though.

  8. So because I’m bored and trying to ignore work, I did some actual linear regression on this. It was a bit of a waste of time, since I had to ignore any data which didn’t have a value for all rows. 12 bits of data makes it hard to say anything is significant, but whatever.

    For the sake of argument, as far as I could tell from just jumping around with models, animation company average is the only really significant predictor for the quality of the anime.

    Stepwise regression gives me a model which also includes series composition average. So a possible model for predicting the MAL rating is:
    1.0287*animation company average + 0.4694*series composition average – 3.9893

    I think I’ve lost control of my life.

  9. My process is as follows:

    >Watch first episode
    >Go to anidb
    >Read roricon’s review of the first episode
    >See that he thinks it’s shit
    >Ignore roricon
    >Watch rest of episodes

    This has yet to let me down.

  10. Here’s the updated list of 2-5 minute anime airing in Winter 2013:

    1. Ai Mai Mi
    2. Hetalia: The Beautiful World
    3. Ishida to Asakura
    4. Jigoku Youchien
    5. Mangirl!
    6. Line Offline Salaryman
    7. Puchimas!
    8. Senyuu
    9. Yama no Susume
    10. gdgd Fairies 2 (15 minutes long)

  11. I don’t really watch many shows as they’re airing. Mainly because of bandwidth issues.

    Also, I tend to have shitty taste, so I don’t pay attention to MAL ratings.


Leave a Comment