Umineko no Naku Koro ni

Review of famous light novel series 'Umineko no Naku Koro ni': it is a highly idiosyncratic, wildly self-indulgent, yet very impressive exploration of all possible locked-room mysteries, knocked for a loop by a tragedy mid-composition, triggering a descent into a deeply harmful endorsement of fantasies & running from painful realities.
reviews⁠, philosophy⁠, anime
2018-07-042021-05-14 finished certainty: highly likely importance: 2 backlinks


One of the anime I enjoy a lot is : a time-loop mystery/fantasy series with hefty helpings of gore and the tried-and-true anime formula of luring the viewer in with comedy and sentimentality, then knocking them for a loop with drama and tragedy. (Other exemplars of this strategy are Fullmetal Alchemist and Clannad.) I was intrigued to learn that the writer of the original ⁠, (Ryukishi07), actually had another mystery/fantasy series which was even longer and in some circles more highly regarded: . There was an anime but I was reassured that it was wretched and the palest reflection of the visual novels. Among anime fans, visual novel fans tend to be the nerdiest and to frequently claim that, while many popular anime were adapted from VNs, the VNs were longer and superior to the adaptations. I’ve seen too many book/movie pairs in Western media where this was also true to simply discount their claims, and I thought that Umineko might be a good way to dip my toes into the VN waters. And if Umineko was crap, I didn’t have to deal with complaints that I had not given it a fair try since everyone agreed the VN was not jut a little bit better than the Umineko anime, but far better. But I kept putting it off since there was no particular reason to read Umineko at any point in time.

In June/July 2013, Tuxedage disappeared from IRC for a few weeks and returned on 2013-07-08, reporting he had been paralyzed for the past 2 weeks by “a dark pit”—finishing the newest version of Umineko with improved graphics—and describing it as “mind-screwy” and saying, “On one hand it’s pure awesome. On the other hand, I kept going”What the fuck did I just read" the entire novel. Even until the end…Umineko is such a screwed up story that I’m still trying to unravel my head around it. I’m watching 10-hour-long videos on YouTube attempting to explain what the heck happened in the novel", even more so than Neon Genesis Evangelion (which I love). I like mind-screwy stuff a lot (like Eco’s , or most of work), I liked Higurashi, a new graphics set sounded like a good time, I was not in the middle of watching any anime (having just finished the short Little Reimu) nor reading any books (I was mostly focused on transcribing Radiance), and Tuxedage’s opinions seem reliable… so why not start now?

I hopped on my usual torrent site to download it (1–4⁠, 5–8). I’d seen screenshots of the original artwork, and they are Touhou-levels of ugly; actually, the first time I happened to see them, I assumed they were someone’s MS Paint-style parodies of the actual Umineko artwork. Thank god for the new artwork! I would have torn out my eyes if I had had to look at the original artwork the entire time. I downloaded it, ran it in WINE without problem, and on 2013-07-15, began playing it…

From a technical perspective, the experience left something to be desired: a small narrow fixed-size window, which could fit relatively little dialogue and made matters more tiring than they had to be. The cramped window and tedium of having to constantly hit the spacebar to advance lines aside (I wish I could just tell it ‘my reading speed is 400WPM, please advance at that rate automatically except for dialogue’; visual novels seem to have the worst of both worlds of novels and anime, where they show as little information as a screen of an anime yet still force you to choose your own pacing without interaction being as easy as a novel where you read an entire page before needing to take any physical action), I’m enjoying it. But none of this was a fatal problem, and the voice acting and limited animation did seem to add something to the experience over flat text. I don’t regret reading it, although I’m not sure I will be tackling any visual novels in the near future. (Tuxedage says that is almost as good as Umineko, and it’s ranked much higher than Umineko on the Visual Novel Database at #1 of all time vs #8; but he cautions it requires a time investment of >30 hours to get the payoff and I’m not sure I have it in me to run another marathon after Umineko, no matter how rewarding it was.)


After ~64 hours of reading time, on 2013-09-21, I finally struggled out the other end.

Tuxedage was right, it was mind-screwy; and awesome, and awful. It was long, intricate, baffling, a gorgeously flawed achievement. Everyone should read it; no one should read it. I still don’t know what to think of it. Is it ridiculous self-indulgent tripe which exposes my own mush-headedness, or the deepest mystery I will ever read? Or does it wander between the 2, tracing the cartography of Ryukishi07’s own life including the unexpected death of his fellow member of 07th Expansion and best friend BT, and, as one person in a 4chan discussion put it:

Umineko is a good editor away from being one of the most interesting VNs in recent memory. The thing is that the story was often times tedious and it got bogged down in the details. I’m pretty sure that if a professional writer streamlined it and cut out some of the heavy handed metaphors and over dramatic declarations of love the thing would be very fucking good.

I find it plausible that Ryukishi07 may never surpass Umineko. (I would analogize Higurashi and Umineko as similar to and Evangelion: the former is, in many respects, a more competently done and solid work, with excellence in all areas from music to artwork to plot to overall conception, without the large flaws and stylistic choices that make many people cringe away from or even hate from the latter; but nevertheless, the lightning struck the latter, those who have consumed the latter may never forget it, and it is easy to predict which will still be debated and watched decades later.)

Renall is, I think, a little too harsh on Umineko when he writes

I’m fascinated by the raw material he [Ryukishi07] works with and the creativity in his expression, but much like the artwork he draws, when I examine it even casually I find it to be full of flaws, corner-cutting, and laziness which detracts from the brilliance that occasionally shows through.

Flaws like characterization issues, redundant writing, excessive description that bogs down scenes that are supposed to be taut and quickly-paced. Corner-cutting like not hiring an artist or editor, changing things to assuage the audience, or not generally acting like the professional that—“doujin work” label or not—he must conduct himself as. Laziness like plot threads and red herrings never addressed and major themes introduced and not adequately developed. I think these things, rather than malice or cynicism, contributed to the schizoid impression of his message and his overall quality that some readers have found.

I like the story, but I do not like the work. I like the style of his character design, but I do not like the implementation in the actual sprites. I like the characters but I do not like their development (or lack thereof). It’s like he’s a miner who found a cache of gorgeous natural rubies, and rather than do the best he can to ensure that they are cut into the most jaw-dropping gemstones anyone has seen in centuries, he tries to do everything himself and produces a bunch of small, amateurishly-cut stones that are generally regarded as a waste of raw material.

I’d love more than anything to see those things shine. But he is not a competent enough craftsman by himself, and until he starts to conduct himself professionally he never will be.

I think is actually a good example of contrast. They became more professional over time and their craftsmanship improved even if their actual content did not. The films are gorgeous and brilliantly-realized, they’re just the nonsense scribblings of a highschooler who played way too much . Ryukishi07 has better ideas, but he isn’t professional at all. It disappoints me far more to see a good idea wasted by laziness than a terrible idea lavished with love and care. A brilliantly-cut piece of cheap artificial crystalware shines more brightly than a scuffed and amateurish ruby, but it’s the latter you wish had the cut of the former, as then you could throw the other one away.

But I have to admit, there is truth to his critique. For every high, there is a low.


I am not going to try to summarize or describe the plot beyond what I have already, and the analysis in another section. I went into Umineko knowing essentially knowing beyond “there are murders and a witch and seagulls”. This meant I was shocked and betrayed by some of the developments, but also that I had no preconceptions or knowledge about any theories or debates. I think probably this is the best way to view works of this type. If you are on the fence, perhaps you should decide now and come back afterwards. I will not be marking spoilers or ROT-13ing passages.

One of the things that hooked me about the plot was how Ryukishi07 could throw in interesting ideas on top of the various mysteries. In episode 8, one of the puzzles for the reader is nothing less than the with a decent explanation of subjective probability. In episode 1, we are given “chessboard thinking”, a memorable metaphor for adversarial reasoning (touching on topics like play and game theory). Not long after that, we see —used correctly and meaningfully! I was impressed. I learned the term too, from Umineko, and that is a useful name to know for expressing an idea about proving negatives and impossibly high burdens of proof. Some of the puzzles tie in with classic attacks or tricks from elsewhere (I came up with a solution to one puzzle due to knowledge of ⁠, and a little poll suggests that cryptography knowledge may help with solving puzzles). Kinzo’s eerie business “luck” turns out to be essentially a ⁠/ lottery⁠. The “magic” system elaborated by Ange and Maria (for characters like the Seven Sisters or Sakutarou) may seem opaque as mud, but when one thinks about daydreams, meditative/psychological practices of creating s (see references), and one looks at the charismatic Christian methods for “hearing God” described in Luhrmann’s When God Talks Back⁠, then it becomes much clearer and much less nonsensical/moronic, and the work that much richer.

One thing I found interesting about the plot was how little blame there is. (“Blaming someone was not the point of my work, and I don’t want anyone to think that it ever was.”) In this respect, it reminds me of Ghibli movies and Hayao Miyazaki. No country is entirely good or evil, it seems, although this has the necessary corollary that your opinions of “good” characters may well sink over the course of the story (Rosa: Worst Mom Ever; Kinzo: Worst Dad Ever). The story is indeed concerned with the “heart” of many characters. I particularly liked Erika Furudo and still catch myself occasionally thinking quotes from her.

But I must speak of the bad as well as good. It’s not that post-episode 5 or so, Umineko starts becoming very silly indeed; I didn’t mind although a lot of people do. (I more minded how the silly “battles” wasted my time with repetitive junk.) Those parts are essentially a very drawn-out epilogue, a recap of the themes, left to clear up the mysteries for those of us who are not geniuses. Erika as pirate-ship captain? Why not.

The real problem is in the core episodes 1–4 or so. When Umineko starts, we are plunged into the plot with no frame story. This is revealed to be a sub-story with a meta-Battler contending with Beatrice the Golden Witch about whether she can force him to acknowledge her and her magic and be fully resurrected into the mortal world. Good, fine. This explains the magic we see on screen just fine, and makes for a very interesting problem: how can our protagonist Battler refute magic, maintaining his disbelief, and eventually turning it into a self-fulfilling prophecy as Beatrice is defeated and her magic collapses? Not too well, it turns out, and things only get worse for Battler as we see ever more magic on screen. Eventually, I wrote that the case against Battler was starting to seem too strong:

I happened to read ANN’s review of the second volume of the Umineko manga and apparently the author remarks that most readers will be convinced that it’s all supernatural and inexplicable (implying that it’ll turn out to be like Higurashi, with naturalistic explanations for (almost) all the events)—which makes me angry since in the dozens of hours I’ve spent reading this, I was pretty sure that we were being shown all the events from third-person omniscient perspective, and one of the narrative rules of is that you don’t lie to the reader, and we see plenty of supernatural events from the omniscient perspective.

…What I’m worried about is that we’re going to be told that the goats, the lightsabers, the parties at the end, the tea parties in another dimension, Rosa being fed her siblings—all the supernatural events we are shown, not told about, in which there’s no need for misleading inference or mistaken assumptions or interpretations—will be somehow waved away like ‘it was all a dream!’ and I’ll be mocked by the author for being so gullible as to believe what I saw on the screen.

If this turned out to be the case, I felt it would be a deep break of the contract between reader and writer: that the reader can trust what he has read or seen. The reader may make an unwarranted inference or interpretation, the reader may choose to ignore framing effects indicating that something is unreliable (for example, if a character read from a diary and it’s visualized on screen, we understand that the diary author may be misleading, misled, the diary a forgery, etc), but otherwise, we can trust what we see. Otherwise, the fiction enterprise falls to pieces. There can be no story where there is no trust. Since fiction is, when you get right down to it, merely particularly “gripping lies”, too much lying can blow a work to smithereens. It’s very dangerous for a writer to take the cliched tactic “it was all a dream” because it threatens the basic integrity of the communication between writer and reader. I thought I could trust Ryukishi07 to not pull a jerkwad move like that, given that in Umineko itself he takes pains to introduce which was a mystery-genre solution to hack writers undermining the integrity of mysteries with impossible and unsatisfying “solutions”.

Then in episode 4 with the revelation that Kinzo had been dead even in episode 1 (which spent hours on Kinzo dialogue and scenes), this is exactly what happened. Umineko even quasi-breaks the fourth wall to mock you for believing anything you had seen on screen. I had no words that day to express my anger: (╯°□°)╯ ┻━┻ …And then, as if that was not enough, we readers get collectively mocked (quite unfairly) some more in the final episode by Ryukishi07.

This ‘treachery of images’ substantially damaged Umineko for me. Ryukishi07, you cannot feed me “it was all a story a character was writing” and expect me to intellectually engage with it as a mystery, when you have violated the most basic contract between reader and writer: I no longer trust you about anything, much less things like Knox’s Decalogue or creating soluble mysteries, and I am certainly not going to invest any mental effort into trying to solve your mysteries before the end. When I was reading , I did spend some time trying to figure out the answer for myself, but that was because the author played it straight with me the entire time, and I had to admit that his solution was fairly done. In retrospect, with the skeleton key to Umineko of Shkanontrice, I do admit that a lot of the mysteries had reasonable solutions and it is a satisfying overall resolution. But I find the work flawed in that no one could figure out the mystery before all was laid bare in the final episodes. In this respect, Umineko was a bad mystery and an object lesson in why one does not violate Van Dine’s second rule: “No willful tricks or deceptions may be placed on the reader other than those played legitimately by the criminal on the detective himself.”

(One might wonder about the riddle/epitaph that is such a MacGuffin, did that bother me? Perhaps surprisingly, not at all: I guessed immediately that it was going to be some stupid kanji-related trick which I didn’t have a snowball’s chance in hell of solving, and I didn’t spare a thought for it. I was maliciously amused when the solution was explained as some ludicrous riddle involving Taiwan of all places and abandoned train stations, and it became clear that Japanese readers hadn’t stood the slightest chance of solving it either and it was the equivalent of a Sherlock Holmes mystery1⁠.)


What’s the real “Single Truth” of what happened on Rokkenjima? As I was fully expecting given the whole magic/willful-blindness theme hammered in again and again, despite the whole book MacGuffin, we never find out for sure; but like Borges’s “Richard Hull, Excellent Intentions⁠, I am sure there is a hidden true mystery underneath the simpler more obvious resolution. Akin to Ange’s ending and ultimately Yasu’s story, I think it’s easily inferred what the true events were. Let’s consider what we know:

  1. events must be consistent with future observations (future witches operate within the factual constraints set down by the past witches):

    • Battler’s body must be alive, probably via the submarine base

    • Eva must be alive, via Kuwadorian (the police would have checked its existence as one of the few things they could check); Eva should also not be the culprit inasmuch as her beloved George is dead

    • the island must be blown up

    • Beatrice must not have killed anyone (at least, I think—IIRC, Virgilia gives this as a red truth in episode 4)

    • they must motivate Eva to never reveal the truth even as the ugliest possible rumors swirl in the media; see Ryukishi07’s post-Umineko interview discussion of what we can infer about the real events (comments starting with “One of the things that you can really know from within the story is that ‘Eva knew something but decided to never to tell Ange about it.’”)

    • they must also motivate Eva to loathe and despise Ange (rather than, say, passionately devote herself to Ange as the last Ushiromiya and all she has left now that her immediate and extended family are all dead)

    • a literary requirement: the culprit must die; in the post-Rokkenjima world, shikata ga nai—there is no longer any guilty party to hunt down or punish. Only Ange’s possibly self-destructive quest for the truth.

    • they must inspire complete revulsion and horror and denial in Ange and Toya/amnesiac-Battler when they learn/remember the truth

    • it must resemble the later episodes penned by Toya/amnesiac-Battler as he worked through his fragmented nightmarish memories; not necessarily in its entirety, but aspects of the truth must be reflected in the various episodes—the setup and motivations in the early episodes penned by Yasu, and facets of what actually happened in Toya’s episodes

      • for example, he failed to figure out the culprit in time and Yasu died on the island
      • in particular, we see his memories involve bloodshed and murder; this rules out eg. any peaceful ‘oops we accidentally set us up the bomb’ scenarios
  2. it cannot match the first 2 episodes since those were penned in advance of the events

  3. the truth should be thematically consistent: just as the truth about Yasu prompts a devastating re-evaluation of the series, so too should the true events at least be shaking.

    One pattern in Umineko is that however bad we think the facts are, they’re worse.

    However ruthless we think Kyrie is early on, she’s much worse (a self-admitted would-be murderess). However abusive we initially think Rosa is of Maria, it turns out to be worse. We think poorly of Kinzo for imprisoning Beatrice Junior and callously disregarding his childrens’ welfare, but the incestuous and explosive truth is much worse. Whatever we assumed the process of getting the Nazi gold entailed, the origins turn out to be much bloodier than we imagined. Whatever we thought of Rudolf as a playboy, we probably did not guess about his infidelity, child-swapping/fake-stillbirth, driving people to suicide, having Yakuza ties, etc. However bad we think Kanon/Shannon’s weird indebtedness to Kinzo and being ‘furniture’ is, when we finally figure out what being ‘furniture’ means, it’s pretty bad. In Ange’s ‘trick’ ending, it’s already clear to us that Amakusa has been tasked with killing her, but the ‘magic’ ending emphasizes this by forcing a reset all the way back to the beginning of Ange’s journey and Ange giving Okonogi her entire fortune—rather than the narratively-simpler strategy of just not going to Rokkenjima or the Sumaderas backing down when there. Yasu is screwed almost from birth, and we learn from Bernkastel that her rejection is almost guaranteed—and in the astronomically unlikely scenario she isn’t rejected and grows up as “Lion”, she dies young anyway.

    This isn’t even a complete list of how bad things are, or how characters perpetuate their hatreds and fears and unhappiness as they abuse each other. Of course we see everyone at their worse, but that’s a pretty lame defense inasmuch as it’s like the murderer pointing out to the court how many days on which he never committed murder. In TvTropes terms, Umineko is a Crapsack World⁠. The Single Truth will be no exception to this. And some character must know it, since as the witches show, there cannot be stories/lies (Magic) without knowing the actual truth (Mystery). A consistent and unfalsifiable set of lies requires knowledge of the truth to avoid getting tripped up or self-contradiction.

With these criteria in mind, there is one story which fits perfectly and violates none of these requirements: the ‘game’ shown in episode 7 to Ange and Lion, in which the Ushiromiya adult siblings collectively solve the Epitaph, find the gold and learn about the bomb, fall out, and Kyrie and Rudolf systematically hunt down and murder everyone.

Episode 7 fits everything: we last know of Battler heading for the chapel right by the tunnel, so the first criteria is met (Battler flees to escape the murders); Eva knows about the bomb and Kuwadorian, so that’s no issue; Beatrice’s epitaph is solved so she never gets started on her killing spree; Eva is motivated to not reveal the truth since she accidentally started the killings (guilt) and cannot tell Ange that her mother was a mass murderer who did not love her and her father was no better and and her big brother was an accomplice (as far as Eva would know); Eva would also be reminded of the murderers who cold-bloodedly killed her husband and son every time she looked at Ange, explaining her hatred of Ange mingled with pride and love in keeping the truth from Ange; while he knows he’s at least innocent, Battler would be little happier about the truth than Eva and likewise doesn’t wish Ange to find out; the use of the tunnels and Kinzo’s ‘test’, among other things, echoes the later episodes; murder is definitely involved; it is completely different from the first 2 episodes; and it’s revolting.

With episode 7 in mind, we can observe further reasons: Bernkastel says that she showed Ange that fragment to gauge her ability to handle the truth, and after Ange fails the test utterly (in a way reminiscent of her reaction to learning the Single Truth in episode 8), strongly encourages her to go after the MacGuffin. Bernkastel being Bernkastel, it’s a safe bet that whatever the Single Truth was, it was very close to the story Bernkastel showed. Further, we are told that Featherine is ‘checking the answers’; given her literally cosmic experience with these stories and the overlap with Toya/amnesiac-Battler, it makes sense that her theory would be very close to the truth and also what we were being shown.

Finally and perhaps most compellingly, from an out-of-universe, it would make little sense for Ryukishi07 to not show us the true story at some point (or something so close to it as to make no difference)! Throughout Umineko, he made a big deal out of characters choosing between “mystery” (the crushing reality of their lives in their crapsack worlds), or “magic” (self-protective delusions); why show us the truth so many times when characters must choose what to believe about smaller stories and incidents, and not give the reader the choice about the overall story? If he did not give us a reasonable chance at making the choice ourselves, it would radically undermine the entire project.


The “magic”/“mystery” theme boils down to Ryukishi07 defending the self-delusion of magic. I strongly disagree with Umineko and its theme. Magic corrupts oneself and frustrates action. The scenarios presented are revolting, and the clear message that we should forget about the past or evil when convenient. (“It is time to let bygones be bygones”, a former Khmer Rouge tells us⁠.) Renall on AnimeSuki calls a spade a spade:

As far as we even know, Beatrice wasn’t guilty of anything, so she doesn’t even have to be forgiven for a murder if there was one. But somebody is. They can’t be forgiven or condemned if that information is unknown.

“Well some things are unknowable!”

Okay, then evil won in Umineko. Evil. Won. And nobody cares because Beatrice and Battler were too selfish, because Eva was too selfish, because Ange was too selfish. Because no one actually cares about anyone but themselves. You ever notice how selfish and possessive so much of the “love” in Umineko is? Hell, Yasu’s love for Battler seems almost hopelessly selfish. The only person who loved unconditionally was Maria and we all see where that got her.

So yeah, I can’t condemn the culprit, but I guess I can condemn all the protagonists. They’re terrible, terrible people. They are indirect agents of the culprit (presuming, again, there was one, and if there wasn’t that should be known) and they ensured his or her victory. There’s your precious “moral”. Money and greed blind a person to the truth. Love blinds people just as hard. That may not be what he’s saying directly, but that’s what the text actually shows: Love and truth are in direct opposition in Ryukishi07’s mind. What a deplorable state of affairs.

Even at the end, I found it incomprehensible how either the author or characters could expect anyone to forgive Yasu, even if they understood, and all she did was try to murder everyone as part of a suicide attempt. How much more so the actual antagonists?

The defense of the “magic” view is also ridiculously bad and completely undercuts the author’s point: he could hardly have made a more convincing case against self-delusion if he tried, and offers us tons of examples where the Litany of Gendlin should have been applied. When Ange, rejected by her schoolmates, dives into magic, she may feel better… at the cost of definitively alienating her classmates, blocking any rapprochement (not that she ever tried), and directly contributing to things like failing her exams (because she was so busy reading Maria’s grimoire and visualizing her friends) and making her classmates angry at her. When Maria utters “~uuu”, my reaction is not “poor Maria!” but “you are bringing this all down on yourself! If you would stop saying ‘uuu’, you would simultaneously stop getting beaten so much, have a chance at friends, and stop being so sad and lonely and despairing”. (I was reading David Chapman’s in-progress novel Buddhism for Vampires recently, and his discussion of “black magic” make the good point that such resentful self-protective magic can often be a crippling crutch.) Ryukishi07 emphasizes this even further by adding scenes where Rosa stonewalls a Child Protective Services agent and orders Maria to avoid other adults (like CPS) and lie and deceive to cover up the abusive truth. The reader cannot know whether Maria would be better off being taken away from Rosa entirely (and this is unknowable, what with it being a fictional work), but it is true that we also have no way of knowing for certain that Maria is best off enabling her mother’s abuse—and certainly neither does Maria; and so her magic winds up being an opiate and one of the direct causes of her continued suffering when there were alternatives available.

I am reminded of : EoE is notoriously opaque and confusing and debatable, but when you read through the script and the deleted scenes and contrast with the end of the TV series, the overall plot is actually fairly simple—Shinji has no reason to live and wants to die/merge with the human collective, then he changes his mind and resolves to live as a flawed separate being. Why the fuss? Because his change of heart makes zero sense. (At least on first view, on its own, without watching a Concurrency edit which fills in the missing pieces.) It is almost completely unmotivated, unexplained, and if Umineko has a toxic crappy family, then the post-apocalyptic setting of the final scene of EoE is even more of a crapsack world (some Eva fans have debated whether, given the Giant Naked Rei corpse, it’s even physically possible for Shinji and Asuka to survive more than a few weeks). This produces tremendous confusion as viewers inherently expect any such change of heart to at least be explained or argued for, not simply happen by fiat in a setting where the choice involves deeply distasteful consequences. But nevertheless, the text is clear, all materials support it, and the primary creator (Hideaki Anno) has said nothing to undermine the prima facie interpretation. However cathartic for the creators it was, and however pure an expression of their personal vision, can we call Umineko or EoE true successes?


The music was as good as the original artwork is bad, which is to say, pretty good; from this, I infer that they accidentally took their entire graphics budget and spent it on music. Below is a list of my favorites; most of these are from the games themselves (provided in OGG or WAV), but “Requiem” is from the discography:


While writing my review, I took 2 geeky detours: estimating how much time, exactly, it had taken me to read through Umineko; and how long is Umineko compared to other works known for being incredibly long like Robert Jordan’s (which I have also read)?

Measuring the actual time spent reading, and comparing the compressed file sizes to normalize the formatting & repetition, Umineko is about as big as Shakespeare but still much smaller than the full Wheel of Time series.

Time investment

How much time did I spend reading through Umineko and then reading about it online? I can get my total time investment from arbtt⁠, by adding the classification rule:

current window $title =~ /.*mineko.*/ ==> tag Umineko,

And then running these queries (I need to do 2 because sometime in July, a log entry was corrupted and as part of the recovery I stored all the previous logs in a separate file):

arbtt-stats --only=Umineko
# Total time per tag
# ____Tag_|_________Time_|_Percentage_
# Umineko |  1d01h00m40s |     100.00
#     WWW |     3h21m20s |      13.42

arbtt-stats --logfile=doc/arbtt/2012-2013.log --only=Umineko
# Total time per tag
# ____Tag_|_________Time_|_Percentage_
# Umineko |  1d13h06m12s |     100.00
#     WWW |       51m52s |       2.33

26 + 25 + 13 = 64 hours. (This excludes any time spent talking about it on IRC or LW.)

Textual length/complexity

Umineko is really long and complex. The total has been estimated at 1.3 million words, there’s something like 63 characters spread over at least 12 discrete storytelling units, with 7 or more different scenarios played out. I was curious how it compared.

A simple word count is not necessarily appropriate: the script is mostly dialogue, which is repetitive compared to your normal novel. In the spirit of the Hutter Prize and Burfoot’s ⁠, a better way might be to approximate the entropy of the various works by examining their compressed sizes. (Another example of this approach is comparing Dubliners to Ulysses⁠.)

The most powerful general-purpose and widely-available lossless compression tool I know of is the xz implementation of (sudo apt-get install xz-utils). It’s not as good as the highly-tuned LZMA variants being used in the Hutter compression contest, but xz is the most powerful widely-available standard form of LZMA I know of and so it’s a better choice than more familiar compression tools like gzip.

The Umineko HTML scripts for episodes 1–6 (minus 7 and 8) are downloaded from the Internet Archive; the Wheel of Time PDFs are downloaded from Libgen (except the last, A Memory of Light, which is apparently unavailable). We extract the text:

elinks -dump \
    > umineko.txt
wget -O shakespeare.txt
emacs -nw shakespeare.txt # manually edit out header, footer, copyright notice in between works
dos2unix *

for file in Robert*.pdf \(Wheel*\)*.pdf; do pdftotext "$file"; done
cat Robert*.txt \(Wheel*\)*.txt > wot.txt && rm *.pdf

With these new files, we can look at their uncompressed sizes:

du -h *
# 5.1M    shakespeare.txt
# 5.4M    umineko.txt
# 21M wot.txt
# 32M total
## 8.8MB when compressed as an XZ tarball
wc *
# 121809   883317 5327372 shakespeare.txt
# 107853   885597 5651767 umineko.txt
# 295640 3928775 21638965 wot.txt
# 525302 5697689 32618104 total

WoT = 3928775; this is consistent with ’s overall estimate of 4056130 total words (if AMoL is ~360k). So far so good.

Now let’s do the compression and see:

for book in *
    echo "$book"

    echo "Gzip (min): "; cat $book | gzip -1 --stdout - | wc --bytes
    echo "Gzip (max): "; cat $book | gzip -9 --stdout - | wc --bytes

    ## commented out because bzip2 insists on printing junk which clutters the output
    # echo "Bzip2 (min): "; cat $book | bzip2 -1 --stdout --compress --quiet | wc --bytes
    # echo "Bzip2 (max): "; cat $book | bzip2 -9 --stdout --compress --quiet | wc --bytes

    echo "LZMA (min): "; cat $book | xz -0 --stdout | wc --bytes
    echo "LZMA (max): "; cat $book | xz -9 --extreme --stdout | wc --bytes


Text Gzip (min) Gzip (max) LZMA (min) LZMA (max)
Shakespeare 2,347,161 1,967,186 2 ,124,932 1 ,549,296
Umineko 2,351,850 1,929,022 2 ,103,496 1 ,427,424
Wheel Of Time 9,796,842 8,216,241 8 ,895,168 5 ,905,940

Umineko vs the compleat Shakespeare: 1427424 / 1549296 = 92%

Versus WoT: 1427424 / 5905940 = 24%

So we see that Umineko contains somewhere around as much entropy as Shakespeare (Shakespeare is shorter, but harder to compress, as one expects), but for sheer bulk, can’t compete with WoT, which makes sense to me: I don’t think I could read WoT in just 70 hours unless I skimmed a lot of it—those books are gargantuan!

  1. Apparently one Japanese who actually was living in Taiwan almost solved it; from a post-Umineko interview:

    KEIYA: I have heard and researched a little about that person who solved the epitaph, it seems it was someone who did a homestay in Taiwan, right‽

    Ryukishi07: It seems like that. I have a faint memory of reading something like “I am overbroad in Taiwan right now”. But, even those this theory was basically the correct answer, it was still following another popular main theory. It’s really like in Umineko itself, isn’t it‽ I thought to myself that this was actually close to the final answer of the riddle that I wanted to give in the main Episodes, but even though there are many followers, there were as many people who wouldn’t believe it and kept searching for alternatives saying, “I don’t buy it, let’s look at it differently!”.