Remove ads

Marked for Death: A Rational Naruto Quest

Voting is open for the next 2 days, 1 hour
Reader mode
faflec said:
Go to the links in my sig.
Click to expand...
Click to shrink...
Shiny.

So I must say, Hazou's practiced motions are pretty grand.

e: Hm. I wonder if Hazou could utilize a custom technique to calculate the exact movements of his opponents, utilizing his high Intelligence in combination with Iron Nerve's innate understanding of a body's motion.

...basically, Poor Man's Sharingan.
 
Last edited:
Remove ads
Omake: Chunin Exam Final
  • Apocrypha
    • Omake: Chunin Exam Final

    "Final match - Uchiha Sasuke versus Kurosawa Hazou!" Kakashi sounded like he was actually taking his capacity as the finals referee seriously, Naruto thought. "Gentlemen, ready?"

    Naruto was surprised how much like an Uchiha the Mist guy looked, black hair and pointy face and all, given how long ago his family had apparently split off. Sasuke had been shocked when the Perv-sage had asked him to confirm that that the other boy could have kids who possessed a version of the Sharingan, and had been looking forward to fighting him ever since.

    The contestants in the arena below dropped into their stances, tensing. Sasuke's eyes flashed red as his Sharingan spun up. Out of the corner of his eye, Naruto saw Kurosawa's jaw go slack and his stance drop fractionally. He noticed he was confused -- not that it was hard, a taijutsu fighter dropping his stance was like an explosive tag going off in terms of things-to-be-confused-about.

    His confusion was rapidly resolved when he saw that Kurosawa's eyes had flipped from blue to red, with three tomoe to mirror Sasuke's.

    "That's... new," Sasuke observed hesitantly, maintaining his own stance.

    "Yes. Yes it is. Kami, do you see this many colors all the time?" Kurosawa responded in a far-away voice.

    "Sensei, should we delay the match and figure this out?" Sasuke called over to Kakashi.

    "No," Kurosawa butted in, "I think I'm figuring this out. Let's give these people a show." His smile looked entirely too friendly for the situation.

    "Fine by me. Begin!" called Kakashi.

    -*-​

    Sasuke noticed he was confused. There was a fist in the side of his face, where his eyes had insisted Kurosawa was throwing a feint, while the knee strike he'd been sure was the real attack had never materialized. He disengaged quickly, trying to focus enough to drown out Naruto's shouted admonitions to "hurry up and hit him" and to reassess his options.

    Kurosawa's taijutsu was, he admitted begrudgingly, impressive, all the more so for its apparent ability to slip through the Sharingan's predictive capability. He wasn't sure if that was just weird closely-related bloodlines interacting, or if the Mist genin had gone ahead and invented a style of taijutsu that broke physics. From sparring with Kakashi, Sasuke had serious doubts that a weapons exchange would be productive, which left ninjutsu.

    "Grand Fireball Jutsu!" He felt the familiar heat build up in his stomach as he pulled chakra out and transformed it into fire, pouring out of his mouth towards Kurosawa. He wasn't feeling especially kind at the moment, he was pretty confident the other boy was good enough not to get cooked, and there were medics on staff and one on Team Uplift (which, if you were to ask him, was a bizarre choice for a team name).

    The fireball vanished suddenly, drawn into a seal attached to a tagged kunai which was headed right for him. He moved to dodge it almost casually, starting to look for where his opponent had vanished to, before realizing that the onrushing air from the vacuum left behind was tearing the seal apart, holy shit, we might all die right here and substituted himself with the furthest of the large rocks scattered around the arena. Immediately he leapt up and away from the shockwave forming behind him, hoping to use his new vantage point to spot Kurosawa.

    From here he could see the hole his opponent had dived into, which meant the ground was now a trap. Fortunately he was headed for one of the arena walls--

    Out of which was suddenly growing a mass of stone tendrils, reaching out for him. He tried valiantly to reorient, dodge through them, push off them, anything to get out and away, but any that would have given him leverage swerved out of the way, and the remainder ensnared him.

    An earth clone emerged from the wall and held a kunai to his neck. "I yield," Sasuke bit out.

    "Ballistic target, predictable target," it said dully. "Even without the Sharingan."

    ...​

    "WHAT WERE YOU THINKING?" Kagome shrieked.

    "I'm sorry, sensei," Hazou mumbled.

    "SORRY? You THREW a SEAL at a FIREBALL! It's only designed to explode safely when destroyed by AIR PRESSURE! If so much as a SPARK had touched it you could have DIED! And there were all those other people in the arena too I guess. And that Uchiha kid! Good job winning, by the way, that stinking eyeball cheater deserved to be taught a lesson. But he woulda died for sure, and then everyone in Konoha would be screaming for us to have our ears stuffed with lupchanzen and marched off to fight on the front lines, or to have us handed over to that snake guy for experiments, or to be forced to eat each others organs in penance, or--"

    "Kagome! Breathing," Inoue-sensei admonished sharply.

    Hazou tried to preempt the rest of Kagome's rant while he was busy recovering his oxygen. "Sensei, I was actually wondering if it might actually be possible for the two of us to have a quick sparring session?"

    She smirked and sat up from her position lounging on one of the park's benches. "You get promoted to chunin and suddenly you think you're hot shit, huh? Okay Mr. Mew, let's see what you've got." She, of course, replicated a plaintive cat's call perfectly.

    Noburi took up a position as impromptu referee. "Ready? Begin!"

    Hazou's eyes spun red as they had in the arena, causing a flinch of collective shock among the rest of the team. He leapt at Inoue, who held her own for a moment before misreading a palm strike that transitioned smoothly to a joint lock-hip throw combo that put her squarely on her ass.

    "Kami, you sure have grown, kid," Inoue-sensei laughed as Hazou helped pull her to her feet. The noncombatants watched in ill-concealed shock as she, after losing a taijutsu spar for the first time any of them could recall, tried to go in for a hair ruffle, which Hazou dodged deftly before reaching over and mussing up her crimson locks with both hands. "Aaaah! Mercy!" she cried, laughing. Hazou ceremoniously withdrew the comb he had kept on him to fix his own hair in the past and handed it over with a small bow.

    "Hazou-sensei, how were you able to activate the Sharingan again? Sasuke isn't here, and you said that when you couldn't use it right after the match you'd assumed that it only worked when you were fighting someone with the Sharingan themselves," Akane queried.

    "I am fighting someone with the Sharingan," he said, wearing Kurosawa Smirk #47: Iiiii know sooomething yooouu dooon't. "Or at least I'm in conflict with him, which seems to be enough. We were playing shogi a little while ago, and he turned his on as part of his capacity training, and suddenly I could do it too. That Nara guy worked it out. So I asked if Sasuke would wait a bit for my turn."
     
    Last edited:
  • Apocrypha
    • @OliWhail As usual, that was great.

    OliWhail said:
    "Kami, you sure have grown, kid," Inoue-sensei laughed as Hazou helped pull her to her feet. The noncombatants watched in ill-concealed shock as she, after losing a taijutsu spar for the first time any of them could recall, tried to go in for a hair ruffle, which Hazou dodged deftly before reaching over and mussing up her crimson locks with both hands. "Aaaah! Mercy!" she cried, laughing. Hazou ceremoniously withdrew the comb he had kept on him to fix his own hair in the past and handed it over with a small bow.
    Click to expand...
    Click to shrink...

    Goddammit that has been my plan for the future for quite some time. When we are finally able to consistently out-roll Inoue, we need to immediately ruffle her hair, and start to call her Mari instead of Inoue-sensei.
     
    OliWhail said:
    Thinking about it, the best graphic would probably be like a Pokemon type-interaction chart, with different colors for "explicitly in the same continuity", "compatible with", and "contradictory"
    Click to expand...
    Click to shrink...

    SOP: Learn How to DataViz

    Wrong! Time for an impromptu datavis lesson.1 ​It's probably worthwhile looking at the spoilers for each study together, and go through the case studies one by one. Also I'm assuming you know when to use bar charts vs pie charts, PDFs vs CDFs, and other basic cases. This is about the general process one should use.

    Now, when building a data visualization our first step must always be to ask ourselves what sort of information we're trying to represent. Try to be as precise as possible. This is the data you have to represent.

    • Prequel/Sequel relationships between omake
      • Directional, One-to-one
    • Happens-Before/Happens-After relationships between omake
      • Implied by Prequel/Sequel Information
      • Directional, One-to-one
    • Continuity Groupings
      • These are sets of omake that have some higher level shared context.
      • Omake can be part of multiple continuities and continuities can themselves nest and overlap.
    • Compatibility Relationships
      • Between pairs of continuity groupings
      • Between pairs of omake
      • Between continuity groupings and omake
      • Non-directional, and One-to-one
    • Attack/Defence relationships between pokemon types
      • It's always a relationship between attack type and the target pokemon's type. These are not actually the same thing, even if they have the same labels.
      • These are always one of the following values: 0x, 1/2x, 1x, 2x
      • Target pokemon can have multiple types, but their response to attacks is always multiplicative. So that you can take information for two pokemon types and combine them easily.
      • If a pokemon has attacks of multiple types, you can combine the information by taking the maximum damage multiplier for each target type. This gives you the type coverage of a pokemon/team given their particular set of attacks.
    • Number of votes a state has in the electoral college
    • Predicted winner of a particular state.
      • A few states have allocate their electoral votes in multiple buckets. Theirs can be split between different candidates.
    • Where should we be putting money to solidify our lead or run up our total?
    • Number of transistors on a processor/mainframe vs. year of introduction of that processor/mainframe.

    And then we must ask what we want a reader to be able to get from the diagram. What are the most important questions a reader must be able to ask the diagram and get answers for? What do we want the user to take away from this diagram? This can change wildly between different diagrams even if they all use the exact same datasets.

    • What order should I read the omake in?
    • If I'm writing an omake, what pieces of lore should I respect?
    • Shit, will my pokemon survive this attack?
    • Do my pokemon offer reasonably good type coverage? As individuals? as a team?
    • Is this pokemon weak to a lot of types?
    • Who is winning each state?
    • Who will win the election as a whole, given this poll/prediction?
    • How solid is each candidate's position? if the election shifted a little is the outcome likely to change?
    • Are there geographic or demographic trends?
    • Is the transistor count actually doubling every two years? maybe it's every 18 months?

    Now it's worth taking a look at the options we have for representing each type of data, and how they work with the various points we want to make. This is just a preliminary view of things, often the existing base visualization types are just a starting point.

    Adjacency Matrix Style Representation


    • Makes it easy to answer "is A compatible with B?"
    • Obscures the ordering of omake, and other grouping relationships.
    Graph Representation


    • Good for getting ordering relationships, but grouping is hard.
    • If you color the nodes for grouping often you'll require so many colors/patterns that it'll be hard to make it colorblind/greyscale friendly.
    Hypergraph Representation


    • Great for continuity groupings, but they're non-directional so they don't really do the ordering/sequel relationships well.
    Direct Attack/Defence Relationships




    • Good for answering the basic form each question, where there's only one type for attack and defence.
    • Makes it difficult to add multiple type relationships together for either attack or defence.
    Type Relationship Graph



    • Great for finding things like type cycles, and self-weaknesses.
    • Hard to represent both weaknesses and immunities at the same time.
    • Gives the illusion that attack type and target pokemon type are the same thing, which they're not.
    Tabular Presentation



    • Makes it very to answer our questions for the single type case.
    • Is somewhat more difficult to get info for multiple type versions of our questions.
    • Is much harder to answer questions about type coverage.
    Matrix Presentation



    • Can be a bit more confusing than the tabular version for the single type version of our questions.
    • Much easier to look at all the multi-type versions of our questions.
    • Take max of multiple rows to get type coverage.
    • Take mult of multiple columns to get actual defense for multiply typed pokemon.
    • Problematically, really only comes into its own when you can easily hide columns and rows you don't care about. That type of interactivity is hard to add given current tools.
    Vote Total Bar



    • Makes it really easy to answer who's winning and by how much.
    • Hides demographic and geographical trends.
    • Hides those weird single electoral votes in some states.
    State Annotated Total Bar



    • Has the same information as other maps.
    • Still makes it easy to tell who's winning and by how much.
    • Provides the information for geographic trends, but it's really hard to see them since you basically need to reconstruct the map.
    • Does split out the single electoral votes.
    • Provides a certain sort of context that none of the other maps really do, in particular tell you where candidates should be focusing their effort in order to solidify a win, or run up their lead.
    Geographic Map



    • Requires labels to get across how many electoral votes each states get.
    • It's not obvious who's winning and who's not.
    • Doesn't do a great job of showing those states where a single EV in a state is allocated differently.
    • Does show geographic trends well.
    • People tend to think about things in terms of compared area, and square milage has jack shit to do with any of the questions we care about.
    Cartogram Scaled by Electoral College Votes
    Comparison of Versions:


    538 Concrete Example:


    • Area on the map now actually captures the thing we care about.
    • Makes it easier to see who's winning.
    • Geographic trends visible.
    • Relative ranking of states difficult to see.
    Linear Plot



    • Makes it hard to see growth factor at a glance
    • Either hides information in the early part of the plot.
    • Exponential, polynomial, and numerous other growth curves look the same on a linear plot.
    Logarithmic Plot



    • Makes the property we want to show a nice clean linear relationship.
    • People are good at linear relationships, it's often a good idea to make sure that the property you care about end up looking like a straight line if it exists.


    Once you've gone through the various options, you can start building your final diagram. It's worth noting that you can combine the types of diagrams you found in the previous step to get something that fits your needs better.




    The cool part of this is realizing we can combine the hypergraph part of our diagram, with the normal directed graph version of our diagram to capture all the information we care about.

    Changes I'd Make:
    • I'd change the undirected lines on the example to directed arrows. Solid for "Direct Prequel/Sequel", dashed for "Happens before/after".
    • Undirected lines (maybe double lines or some other form of marking) between different continuity groups would indicate compatibility between groups of omake.
    Interactivity Additions:
    • The big one here would be to make sure that links were clickable, there's not actually much else. This one is pretty capable on its own.
    Notes:
    • Pay attention to how the flow of reading always follows from left to right, in in-story chronological order. As long as there is a chronological order, that sort of arrangement is possible.
    • Seriously, you can start combining these different diagram types to get the properties you want. It's not often as elegant as this, but it's often a really good idea when you've got a bunch of things going on at once.



    The reason we ended up with the matrix representation is that it provides the best balance between answering the single-type version of our questions and the multi-type version of our questions. If instead we wanted to find type cycles, then we should go with the type relationship graph. If only the single type version was important we should probably go with the tabular representation.

    Changes I'd Make:
    • This needs a key. It might seem obvious what each of the symbols mean, but it isn't to everyone. Is "+" good for the attacker? the defender?
    • Likewise I'd be a bit more clear on the labels. My version would have "Attack Type" and "Defending Pokemon Type".
    Interactivity Additions:
    • It would be nice to be able to click a bunch of columns or rows and have the corresponding type coverage or aggregate defense pop up.
    Notes:
    • Look at how the numbers were dropped out earlier matrix example, instead they doubled down on the colors and symbols. The symbols are easier to read, and it's all much cleaner. This is even good for the colorblind.
    • Also note how the 1x squares are just gray boxes. As the common case they're not important, and are therefore deemphasized. This is generally good practise.



    Cartograms are incredibly powerful tools when you want to talk about how different states/countries/pieces of land/etc compare to each other in ways that aren't just about land area. The main problem is that they're hard to make, and often require a more artistic touch to do properly.

    Changes I'd Make:
    • The colors should match a bit better, and in particular the top bar should be a straight version of state annotated total bar
      with the state labels hidden. You just get this because I'm lazy.
    Interactivity Additions:
    • Holding your mouse over a state or the total bar should highlight the state that you're hovering over. This highlighting should appear on both the bar and map. Admittedly without this it's hard to answer the tactics questions we care about.
    Notes:
    • Again, another example of just combining multiple types of graph for better understanding. In this case it's just having them next to each other, but it still is useful.
    • You often aren't actually limited to a single image. The 538 Election forecast (backup image) is a master class in representing the same data in a number of different ways depending on your aims.



    This is probably the most straightforward example in this post. Basically, people understand linear relationships so try to use axes that make important trends look like straight lines. In addition, this is a good example of how to annotate your diagrams to get information across.

    Changes I'd Make:
    • I generally like this one, I'd make a few changes to make it prettier, but that's the lot of it.
    Interactivity Additions:
    • This already pretty good. If you have the luxury of interactivity, I'd hide the processor labels and just show them when you hover over them.
    Notes:
    • You can be explicit with the takeaways. That pair of lines showing trends for both the 18 and 24 month doubling makes it obvious what is actually happening. When you have a bunch of hypothesis, showing what each hypothesis would result in is generally good idea. You might have to balance this with clutter, but it's easy enough to only show some of them at a time.

    When you've got your base diagram you can then focus on polish. Here's my usual checklist of stuff to pay attention when polishing my imagery.

    • Remove everything that can be removed: Gridlines, colors, text, etc..
      • For each little component ask yourself: Is my diagram worse without this thing? If not, remove it.
    • Make sure you don't fuck over the colorblind folks. Your diagram really should work when converted to greyscale.
      • This can be entirely selfish in motivation, you'll have to present with buggy projectors or print with a B/W printer.
      • Don't use cross hatching, just don't.
    • Don't be afraid to make multiple versions of a diagram for a presentation. It's often a really good idea to have a single diagram with different sections highlighted as you want to emphasize different points.
    • When looking at particular option for a diagram, think about how it could be misinterpreted.
      • Take a look at the final version of the pokemon type diagram. Its symbols are ambiguous without a key. It's really worth your while to sit down and try to to figure out how people will misread your diagram. Also ask others to do the same if possible, you're more likely to miss things on the diagrams you make.
      • Assume people don't read the caption. People don't.
      • These misunderstandings can usually be quickly corrected with a key or a label. Don't worry too much about people who don't read a few short words on the image itself.
    • Don't forget units on your axis labels.
    • Include a key unless the meaning is blindingly obvious. (Note: It's never blindingly obvious)
    • If you're making something 3D, you better have an incredibly good reason. It's a bad idea 99% of the time.

    Finally, here's some resources for learning more, seeing datavis done well or badly, and tools for making graphs and images.

    • Information is Beautiful: Lots of pretty visualizations, often they prefer being artistic over being informative but still worth looking at.
    • Viz.WTF: Visualizations gone wrong. These are examples of how to not make diagrams, but looking at failures is both fun and a great learning experience.
    • Edward Tufte's Books and Courses: The master of getting information across in images. If you can, read his books, they're full of case studies and examples for a huge variety of different applications.
    • Darkhorse Analytics: Their blog has a number of good tips on exact steps to take when making your designs prettier. It's a great resources for learning how to do that final polishing step.
    I'm a bit weird, personally I prefer making my diagrams programmatically and using FOSS whenever possible. For the most part the right answer is to use the tools you know and have access to, this is just what I tend to use.
    • LaTeX2​ and TikZ: The former is a document typesetting system that is the standard in mathematics and computer science. It's honestly the only good way to typeset math and other complex academic documents. The latter is a macro package (basically a library) for making images within LaTeX, and is one of the best ways to get complex static 2d diagrams with complex math. It's really the only programmatic diagramming tool I've encountered that lets me independently describe and style diagrams.
    • Inkscape: An open vector drawing tool. The interface has a bit of a harsh learning curve, but it's very powerful.
    • Gimp: Open source photoshop, good at what it does.
    • GraphViz: Tool for making connectivity graph like diagrams, can often automatically lay them out for you automagically.
    • matplotlib: Python library for creating various plots.
    • gnuplot: Command line tool for generating plots.

    Also, if anyone asks why this is posted here. It's the SOP we link to when Hazou is to make diagrams in story for any reason. He obviously won't use the same terminology, but he should use the same thought process. This should get us a solid +2 to all rolls involving creating diagrams, right @eaglejarl? :whistle:

    [1]: @OliWhail, you're an academic now. This is genuinely really important for you as you write papers. Pictures really are worth a thousand words, but only if you work to make them useful. Not to mention this is a lesson I really enjoy giving. I hope you'll forgive me any condescension in the tone, it's unintentional and hard to regulate over text. Also I plan to show this to other people when I need to, so it's written for a more general audience who probably have less background than you.

    [2]: The 'Tex' part is pronounced "tech", and full thing is either "lay-tech" or "lah-tech" depending on who you ask. If you pronounce it "lay-tex" you are a heathen who deserves to burn for your sins.
     
    Last edited:
    I just had a wonderful, terrible idea for an omake. I'm not sure whether I have the wherewithal to write it out myself, so I'm putting the summary in a spoiler.
    Imagine how the team would have reacted had Mari died in that cave at VHitM from bleeding out just after they broke through the walls. Somehow Takahashi convinces Keiko to go through with the summoning scroll thing anyway, and she impresses Pangolin boss thoroughly from the word "go". She comes out and demands the death of every member of the clans responsible (which Takahashi manages to talk her down to just those responsible and the clan heads) as their Summon-Jesus.

    Meanwhile, Kagome and Noburi are both blaming themselves. Hazou demands Noburi keep manually cycling chakra through her body and repair the tissue in spite of her being dead, tying off her leg. They fail.

    That said, you are an insanely productive writer, OliWhail. Geez! Great job on this!

    e: Changed the last sentence of my omake-summary.
     
    Last edited:
    Jello_Raptor said:
    SOP: Learn How to DataViz

    Wrong! Time for an impromptu datavis lesson.1 ​It's probably worthwhile looking at the spoilers for each study together, and go through the case studies one by one. Also I'm assuming you know when to use bar charts vs pie charts, PDFs vs CDFs, and other basic cases. This is a

    Now, when building a data visualization our first step must always be to ask ourselves what sort of information we're trying to represent. Try to be as precise as possible. This is the data you have to represent.

    • Prequel/Sequel relationships between omake
      • Directional, One-to-one
    • Happens-Before/Happens-After relationships between omake
      • Implied by Prequel/Sequel Information
      • Directional, One-to-one
    • Continuity Groupings
      • These are sets of omake that have some higher level shared context.
      • Omake can be part of multiple continuities and continuities can themselves nest and overlap.
    • Compatibility Relationships
      • Between pairs of continuity groupings
      • Between pairs of omake
      • Between continuity groupings and omake
      • Non-directional, and One-to-one
    • Attack/Defence relationships between pokemon types
      • It's always a relationship between attack type and the target pokemon's type. These are not actually the same thing, even if they have the same labels.
      • These are always one of the following values: 0x, 1/2x, 1x, 2x
      • Target pokemon can have multiple types, but their response to attacks is always multiplicative. So that you can take information for two pokemon types and combine them easily.
      • If a pokemon has attacks of multiple types, you can combine the information by taking the maximum damage multiplier for each target type. This gives you the type coverage of a pokemon/team given their particular set of attacks.
    • Number of votes a state has in the electoral college
    • Predicted winner of a particular state.
      • A few states have allocate their electoral votes in multiple buckets. Theirs can be split between different candidates.
    • Where should we be putting money to solidify our lead or run up our total?
    • Number of transistors on a processor/mainframe vs. year of introduction of that processor/mainframe.

    And then we must ask what we want a reader to be able to get from the diagram. What are the most important questions a reader must be able to ask the diagram and get answers for? What do we want the user to take away from this diagram? This can change wildly between different diagrams even if they all use the exact same datasets.

    • What order should I read the omake in?
    • If I'm writing an omake, what pieces of lore should I respect?
    • Shit, will my pokemon survive this attack?
    • Do my pokemon offer reasonably good type coverage? As individuals? as a team?
    • Is this pokemon weak to a lot of types?
    • Who is winning each state?
    • Who will win the election as a whole, given this poll/prediction?
    • How solid is each candidate's position? if the election shifted a little is the outcome likely to change?
    • Are there geographic or demographic trends?
    • Is the transistor count actually doubling every two years? maybe it's every 18 months?

    Now it's worth taking a look at the options we have for representing each type of data, and how they work with the various points we want to make. This is just a preliminary view of things, often the existing base visualization types are just a starting point.

    Adjacency Matrix Style Representation


    • Makes it easy to answer "is A compatible with B?"
    • Obscures the ordering of omake, and other grouping relationships.
    Graph Representation


    • Good for getting ordering relationships, but grouping is hard.
    • If you color the nodes for grouping often you'll require so many colors/patterns that it'll be hard to make it colorblind/greyscale friendly.
    Hypergraph Representation


    • Great for continuity groupings, but they're non-directional so they don't really do the ordering/sequel relationships well.
    Direct Attack/Defence Relationships




    • Good for answering the basic form each question, where there's only one type for attack and defence.
    • Makes it difficult to add multiple type relationships together for either attack or defence.
    Type Relationship Graph



    • Great for finding things like type cycles, and self-weaknesses.
    • Hard to represent both weaknesses and immunities at the same time.
    • Gives the illusion that attack type and target pokemon type are the same thing, which they're not.
    Tabular Presentation



    • Makes it very to answer our questions for the single type case.
    • Is somewhat more difficult to get info for multiple type versions of our questions.
    • Is much harder to answer questions about type coverage.
    Matrix Presentation



    • Can be a bit more confusing than the tabular version for the single type version of our questions.
    • Much easier to look at all the multi-type versions of our questions.
    • Take max of multiple rows to get type coverage.
    • Take mult of multiple columns to get actual defense for multiply typed pokemon.
    • Problematically, really only comes into its own when you can easily hide columns and rows you don't care about. That type of interactivity is hard to add given current tools.
    Vote Total Bar



    • Makes it really easy to answer who's winning and by how much.
    • Hides demographic and geographical trends.
    • Hides those weird single electoral votes in some states.
    State Annotated Total Bar



    • Has the same information as other maps.
    • Still makes it easy to tell who's winning and by how much.
    • Provides the information for geographic trends, but it's really hard to see them since you basically need to reconstruct the map.
    • Does split out the single electoral votes.
    • Provides a certain sort of context that none of the other maps really do, in particular tell you where candidates should be focusing their effort in order to solidify a win, or run up their lead.
    Geographic Map



    • Requires labels to get across how many electoral votes each states get.
    • It's not obvious who's winning and who's not.
    • Doesn't do a great job of showing those states where a single EV in a state is allocated differently.
    • Does show geographic trends well.
    • People tend to think about things in terms of compared area, and square milage has jack shit to do with any of the questions we care about.
    Cartogram Scaled by Electoral College Votes
    Comparison of Versions:


    538 Concrete Example:


    • Area on the map now actually captures the thing we care about.
    • Makes it easier to see who's winning.
    • Geographic trends visible.
    • Relative ranking of states difficult to see.
    Linear Plot



    • Makes it hard to see growth factor at a glance
    • Either hides information in the early part of the plot.
    • Exponential, polynomial, and numerous other growth curves look the same on a linear plot.
    Logarithmic Plot



    • Makes the property we want to show a nice clean linear relationship.
    • People are good at linear relationships, it's often a good idea to make sure that the property you care about end up looking like a straight line if it exists.


    Once you've gone through the various options, you can start building your final diagram. It's worth noting that you can combine the types of diagrams you found in the previous step to get something that fits your needs better.




    The cool part of this is realizing we can combine the hypergraph part of our diagram, with the normal directed graph version of our diagram to capture all the information we care about.

    Changes I'd Make:
    • I'd change the undirected lines on the example to directed arrows. Solid for "Direct Prequel/Sequel", dashed for "Happens before/after".
    • Undirected lines (maybe double lines or some other form of marking) between different continuity groups would indicate compatibility between groups of omake.
    Interactivity Additions:
    • The big one here would be to make sure that links were clickable, there's not actually much else. This one is pretty capable on its own.
    Notes:
    • Pay attention to how the flow of reading always follows from left to right, in in-story chronological order. As long as there is a chronological order, that sort of arrangement is possible.
    • Seriously, you can start combining these different diagram types to get the properties you want. It's not often as elegant as this, but it's often a really good idea when you've got a bunch of things going on at once.



    The reason we ended up with the matrix representation is that it provides the best balance between answering the single-type version of our questions and the multi-type version of our questions. If instead we wanted to find type cycles, then we should go with the type relationship graph. If only the single type version was important we should probably go with the tabular representation.

    Changes I'd Make:
    • This needs a key. It might seem obvious what each of the symbols mean, but it isn't to everyone. Is "+" good for the attacker? the defender?
    • Likewise I'd be a bit more clear on the labels. My version would have "Attack Type" and "Defending Pokemon Type".
    Interactivity Additions:
    • It would be nice to be able to click a bunch of columns or rows and have the corresponding type coverage or aggregate defense pop up.
    Notes:
    • Look at how the numbers were dropped out earlier matrix example, instead they doubled down on the colors and symbols. The symbols are easier to read, and it's all much cleaner. This is even good for the colorblind.
    • Also note how the 1x squares are just gray boxes. As the common case they're not important, and are therefore deemphasized. This is generally good practise.



    Cartograms are incredibly powerful tools when you want to talk about how different states/countries/pieces of land/etc compare to each other in ways that aren't just about lange area. The main problem is that they're hard to make, and often require a more artistic touch to do properly.

    Changes I'd Make:
    • The colors should match a bit better, and in particular the top bar should be a straight version of state annotated total bar
      with the state labels hidden. You just get this because I'm lazy.
    Interactivity Additions:
    • Holding your move over a state or the total bar should highlight the state that you're hovering over. This highlighting should appear on both the bar and map. Admittedly without this it's hard to answer the tactics questions we care about.
    Notes:
    • Again, another example of just combining multiple types of graph for better understanding. In this case it's just having them next to each other, but it still is useful.
    • You often aren't actually limited to a single image. The 538 Election forecast (backup image) is a master class in representing the same data in a number of different ways depending on your aims.



    This is probably the most straightforward example in this post. Basically, people understand linear relationships so try to use axes that make important trends look like straight lines. In addition, this is a good example of how to annotate your diagrams to get information across.

    Changes I'd Make:
    • I generally like this one, I'd make a few changes to make it prettier, but that's the lot of it.
    Interactivity Additions:
    • This already pretty good. If you have the luxury of interactivity, I'd hide the processor labels and just show them when you hover over them.
    Notes:
    • You can be explicit with the takeaways. That pair of lines showing trends for both the 18 and 24 month doubling makes it obvious what is actually happening. When you have a bunch of hypothesis, showing what each hypothesis would result in is generally good idea. You might have to balance this with clutter, but it's easy enough only show some of them at a time.

    When you've got your base diagram you can then focus on polish. Here's my usual checklist of stuff to pay attention when polishing my imagery.

    • Remove everything that can be removed: Gridlines, colors, text, etc..
      • For each little component ask yourself: Is my diagram worse without this thing? If not, remove it.
    • Make sure you don't fuck over the colorblind blind folks. Your diagram really should work when converted to greyscale.
      • This can be entirely selfish in motivation, you'll have to present with buggy projectors or print with a B/W printer.
      • Don't use cross hatching, just don't.
    • Don't be afraid to make multiple versions of a diagram for a presentation. It's often a really good idea to have a single diagram with different sections highlighted as you want to emphasize different points.
    • When looking at particular option for a diagram, think about how it could be misinterpreted.
      • Take a look at the final version of the pokemon type diagram. Its symbols are ambiguous without a key. It's really worth your while to sit down and try to to figure out how people will misread your diagram. Also ask others to do the same if possible, you're more likely to miss things on the diagrams you make.
      • Assume people don't read the caption. People don't.
      • These misunderstandings can usually be quickly corrected with a key or a label. Don't worry too much about people who don't read a few short words on the image itself.
    • Don't forget units on your axis labels.
    • Include a key unless the meaning is blindingly obvious. (Note: It's never blindingly obvious)
    • If you're making something 3D, you better have an incredibly good reason. It's a bad idea 99% of the time.

    Finally, here's some resources for learning more, seeing datavis done well or badly, and tools for making graphs and images.

    • Information is Beautiful: Lots of pretty visualizations, often they prefer being artistic over being informative but still worth looking at.
    • Viz.WTF: Visualizations gone wrong. These are examples of how to not make diagrams, but looking at failures is both fun and a great learning experience.
    • Edward Tufte's Books and Courses: The master of getting information across in images. If you can, read his books, they're full of case studies and examples for a huge variety of different applications.
    • Darkhorse Analytics: Their blog has a number of good tips on exact steps to take when making your designs prettier. It's a great resources for learning how to do that final polishing step.
    I'm a bit weird, personally I prefer making my diagrams programmatically and using FOSS whenever possible. For the most part the right answer is to use the tools you know and have access to, this is just what I tend to use.
    • LaTeX2​ and TikZ: The former is a document typesetting system that is the standard in mathematics and computer science. It's honestly the only good way to typeset math and other complex academic documents. The latter is a macro package (basically a library) for making images within LaTeX, and is one of the best ways to get complex static 2d diagrams with complex math. It's really the only programmatic diagramming tool I've encountered that lets me independently describe and style diagrams.
    • Inkscape: An open vector drawing tool. The interface has a bit of a harsh learning curve, but it's very powerful.
    • Gimp: Open source photoshop, good at what it does.
    • GraphViz: Tool for making connectivity graph like diagrams, can often automatically lay them out for you automagically.
    • matplotlib: Python library for creating various plots.
    • gnuplot: Command line tool for generating plots.

    Also, if anyone asks why this is posted here. It's the SOP we link to when Hazou is to make diagrams in story for any reason. He obviously won't use the same terminology, but he should use the same thought process. This should get us a solid +2 to all rolls involving creating diagrams, right @eaglejarl? :whistle:

    [1]: @OliWhail, you're an academic now. This is genuinely really important for you as you write papers. Pictures really are worth a thousand words, but only if you work to make them useful. Not to mention this is a lesson I really enjoy giving. I hope you'll forgive me any condescension in the tone, it's unintentional and hard to regulate over text. Also I plan to show this to other people when I need to, so it's written for a more general audience who probably have less background than you.

    [2]: The 'Tex' part is pronounced "tech", and full thing is either "lay-tech" or "lah-tech" depending on who you ask. If you pronounce it "lay-tex" you are a heathen who deserves to burn for your sins.
    Click to expand...
    Click to shrink...
    Omake where Hazou interrupts a Kage summit to chastise them for their use of the wrong diagrams?
    ChronOblivion said:
    @Cariyaga Ugh that's devastating. Personally, I'd cut out that last sentence.
    Click to expand...
    Click to shrink...
    Yeah I wasn't sure about that either, but I saw OliWhail posted an omake and was like "oooh! gonna stop revising this now." e: On thinking about it I'd definitely cut out that last sentence. It's all the better if Hazou does everything he can and still fails.
     
    Last edited:
    Jello_Raptor said:
    Great for continuity groupings, but they're non-directional so they don't really do the ordering/sequel relationships well.
    Click to expand...
    Click to shrink...
    Couldn't we just--
    Jello_Raptor said:
    The cool part of this is realizing we can combine the hypergraph part of our diagram, with the normal directed graph version of our diagram to capture all the information we care about.
    Click to expand...
    Click to shrink...
    Yeah, that :D

    I love it, this is super helpful, both for stuff I might do in the thread and for my entrance into academia :)

    Cariyaga said:
    Omake where Hazou interrupts a Kage summit to chastise them for their use of the wrong diagrams?
    Click to expand...
    Click to shrink...
    DAMMIT PEOPLE I HAVE TESTS TO STUDY FOR STOP INSPIRING ME :mad: @Jello_Raptor are you willing to handle this one?
     
    Last edited:
    You know what we need? A list of omake ideas for adoption. ...Wait, that'd probably kill OliWhail's grades...
     
    OliWhail said:
    So like, I have to think this is clearly at their second, much more secret Kage summit, where they all come in and dispel and punch each other to check for henges but somehow he slips in anyway. Maybe he's henged into the table? :D
    Click to expand...
    Click to shrink...

    :o :evil: :whistle:

    OliWhail said:
    . Sasuke's eyes flashed red as his Sharingan spun up.
    Click to expand...
    Click to shrink...

    I'm just imagining Sasuke's sharingan spooling up, instead of spinning up, and suddenly he's able to FTL jump interstellar distances. Honestly, given the lack of navigation it would one of the less OP canon sharingan uses.

    OliWhail said:
    Kurosawa Smirk #47: Iiiii know sooomething yooouu dooon't.
    Click to expand...
    Click to shrink...

    Of course this is thing. There's no way this isn't a thing.

    Also someone needs to make a canon + omake list of all the Kurasawa standard motions.
     
    Reader mode
    You must login or register to reply here.
    Voting is open for the next 2 days, 1 hour
    Remove ads
    Back
    Top