Greetings, glorious adventurers! If you're joining in our Alpha One spot testing, please follow the steps here to see all the latest test info on our forums and Discord!

Information Compression Mechanisms

ImmortalmageImmortalmage Member, Alpha One, Adventurer
edited December 2020 in General Discussion
TL;DR: A post for people who like game theory. Consider thee fairly warned.

Defining Information Compression

A high resolution representation of the world is one that encompasses all the details, regardless of relevance or utility. You perceive many things with your 5 senses but video games concern only audiovisual experiences and I'll keep this posts concerns there. Visual systems are high bandwidth and people have no issues with audio compliments to that visual experience because we're hard wired to use those systems together.

Preface

This post comes from the perspective of a DND GM with some MMO experience. I've seen some Ashes of Creation content and it's piqued my interest. I won't concern data compression. I consider Information Filtration interchangable with Information Compression and unreasonably incomprehensively complex interchangable with infinitely complex. I make no disclaimers that these processes uncover good game design, they're purely academic feature exploration.

Discovering Linear Information Compression

Any time there is a lower complexity version of something else, some process has compressed information to arrive at the new state. Minimaps are very high, precise representations of a game world that is too much for people to take in all at once. The details that are of consequence are hidden among things that aren't of concern at every moment. The minimap strips away much of the information about the world and gives an uncomplicated view. It strips away the y axis, which is an exponential reduction in complexity. It strips away the parts of the world you're not at, which is useful because there's way more places you aren't than where you are. The last choice is what additional pieces of information should be brought from the high resolution representation to the foreground of the uncomplicated assistive view. Herbs in games that are nearby are often shown here because it tells you exactly where something you care about is on the x and z axis. Arriving may give you a hill climbing problem but that's where people stop looking at the minimap and bring their gaze back to the actual game world where all the details actually are.

This is a linear expression because you can map one thing in the ground truth to one thing in the lower resolution representation view of the ground truth. The compression is related to the fact that many things are removed and only that which remains builds the new view. The more things that aren't shown the more useful that minimap will be to find the things that are on it. The ground truth herb has features like it's 3d model & texture, its x, y, and z, tint, if it's being moused over, when it is, perceptability, etc. Some of those features are relevant based on how they relate to something else. For example, you only care about where the herb is because you care if it is close, and technically speaking it being close is the difference between where you are and where it is. You only care when the herb is because it's only a useful herb if it's there when you're there too. After broken down into its individual components, a compressed view of it on a minimap can be rebuilt in a way that is useful from the perspective of players and game designers alike. Building it back up might reveal some things that were already discovered, such as using a mask to crop out everywhere that's not immediately surrounding the player, because it's not useful to know about terrain half way around the world. Besides the obvious, there's some new features that can be built out into game mechanics. Knowing when a herb is is a nontraditional display of information. Knowing if a herb was recently picked or is almost ripe is two things that would be known by a ground truth and can be pushed to the players foreground with a small representation of that information. With this process of breaking feature down and then excluding many features in a new view, information can be usefully compressed in a way that helps players not only understand the ground truth better but also achieve goals while interacting with the game world.

Not all information that can be compressed should be. With any abstraction, you lose the clarity of the ground truth in exchange for utility. Searching out herbs is a Traveling Salesman Problem (https://en.wikipedia.org/wiki/Travelling_salesman_problem). Computers can do the work to find optimal solutions to the path that can be taken to get the most value out of shortest route between each node. That's a very useful collapse of all possible routes into a view of only the most efficient route. It's not obvious that such information should be filtered for players.

I won't repeat what's said about the interplay between filtration as a CPU optimization parameter on clients and servers, but I will link to Significance Buckets from Fortnite at GDC: https://youtu.be/KHWquMYtji0?t=331

Nonlinearity and Iconification

/_\
|__|

The above is a house. Here's a sword: -|=>. But it's not. It's kind of weird that you can sort of play along that the 6 ASCII characters above is actually a house. The iconification of something is horrendously impossible to get computers to do, but this computationally heavy operation is done automatically and instantly by people. As game designers, taking advantage of this is typically done unconsciously because that efficiency is innate in themselves as well. The nonlinearity of this comes from the fact that not every feature of the ground truth is mappable to one feature of the icon.

This is a house too: https://en.wikipedia.org/wiki/House`#/media/File:Bhutanese_Farmhouse_Soe_Yaksa.jpg

Here is a birdhouse, which is categorically also a house: https://en.wikipedia.org/wiki/House#/media/File:Vogelhaus_Modell_Eduard_4.JPG

These are definitely houses. The icon of the house is a metaset of the above and all other houses. If anything, the icon is more real than any one of the items of the subset of houses despite that the houses used to create the icon are actually real. The abstraction being more real than the thing it represents is not a new idea but functions as if true so I take it as true as such.

There's already Icons in Ashes of Creation. Abilities with ability icons are a nonlinear representation of what pressing that button will do. That icon represents a set of consequences. You'll lose some resource and you'll gain some advantage. There's no necessary correlation between the arrangement of pixels and the resources lost and advantage gained. Item icons are similarly mainstays. In some games I've played, one icon doesn't necessarily only represent one thing. One sword icon can actually represent many swords models, and sword models might represent many sword items because it's easier to create numerical differences in a stat block than it is to create a graphics asset. There will tend to be an outpacing of one unless they are all equally time consuming to create while simultaneously being equally important. You can only rotate or tint an asset so many times because people dislike the reuse because the raw icon lost its unique impact.

Dangerous Emojilization

The race to the bottom of representational states is occurring in my estimation sadly right before our eyes. The prevelance of emoji's is increasing everywhere as they're in practice micro-icons in that they increase the understanding of a message even by those that are English-As-A-Second-Language and other incompetant or functionally illiterate readers. This lowering of the bar means that everyone above the new bar placement is able to understand, but is similarly a sacrifice of the precision of what the sentence could represent for an abstraction that is more accessible along one dimension of analysis.

Iconification of things in game design has similar consequences to the proliferation of iconified representations in the real world. Although there is a place for such processes, there are times when you shouldn't sacrifice the real for the abstract despite the allure of the utility thereof.

- Immortalmage
B)
7U1GDCU.png

Comments

  • You're supposed to put a "TL:DR" in there.....
    This link may help you: https://ashesofcreation.wiki/
  • KhronusKhronus Member, Alpha One, Adventurer
    What is the point of this? I stopped reading after you explained that video games don't use touch, smell and taste.
  • ImmortalmageImmortalmage Member, Alpha One, Adventurer
    daveywavey wrote: »
    You're supposed to put a "TL:DR" in there.....

    A summary is itself the result of the compression of information. You reduce the components to it's essential elements leaving only the high impact portions behind. If I had a summary there and it was intriguing, it could focus people back at the detailed post for more precision in the concepts I'm trying to write about.

    Information Compression Part 2 of 77

    One thing that makes the compression of icons easier is that the process is reversible. Looking at the sets of houses means you can do two things.

    Preface: Creating Nonlinear Icons Using Sets

    The first thing that can be done is commonality analysis. Between things in a set, you can average and compare things to a proposed icon until an obvious feature set reveals itself. Below are some houses and a few tags to describe them.

    https://en.wikipedia.org/wiki/House#/media/File:Kamena_kuca_u_Pokreveniku.jpg

    4 Rectangular Walls also arranged rectangularly, Triangular Roof, Stone Construction, Ground Floor, Enclosed Space, Immovable Foundation

    https://en.wikipedia.org/wiki/House#/media/File:Cambo_169.jpg

    6 Rectangular Walls, Elevated Floor, Wooden Construction, Windowed Enclosed Space, Immovable Foundation, Two Peaked Roof

    https://en.wikipedia.org/wiki/House#/media/File:Brgule_006.jpg

    4 Square Walls arranged in a square, Windowed Enclosed Space, Brick Construction, Elevated Floor, Ground Floor, Oddly Peaked Roof, Balconies, Immovable

    https://en.wikipedia.org/wiki/Gingerbread_house#/media/File:Gingerbread_house_with_gumdrops.jpg

    4 Differently Shaped Walls, Enclosed Space, Peaked Roof, Gingerbread Construction

    If you prune all of the features that aren't shared across all examples, you get:

    At least 4 walls enclosing a space with at least one floor and a roof.

    If you take features by plurality instead of by complete commonality, you get:

    A walled enclosure with a peaked roof.

    /_\
    |__|

    It's easy to see how this uncompresses to the concept of a house now that we've gone through the process in one direction to create the icon. This icon packs all the information of the concept of a house into itself.

    Moving On To Game Design: Creating Nonlinear Things Using Sets

    Typically, levels are a permanent way to increase a character's stats. Gear is a semi-permanent item that you can wear that alters your character's properties, often regarding stats. Consumables are items that provide temporary alterations to your characters properties.

    These things all have sets of stats. Thinking in terms of sets, you can boil items down to individual features and then explore altering those features before building them back up.

    Taking a look at the definition of temporary, that is so commonly time-limited effects that it's what most people probably think of when reading this. Temporary in games really just means a triggered but limited thing - IE, a bounded duration. That trigger can be anything. Thinking that way, it's easy to just think up consumables with examples that aren't just time limited consumption devices.

    * The daytime or nighttime is a bounded duration - From sunrise to sunset or vice versa.

    * If you're a werewolf, your properties change at the start of a full moon and end afterwards.

    * Entropic decay, IE The construction of something that breaks down.

    * Geographically constrained / geofenced stat adjustment.

    * Concurrent events or randomness (Randomness like critical chance, concurrent things like critting multiple times in a row)

    * While stat relationships hold true (More Constitution than Strength, exactly equal Wisdom and Intellect, more than X health but less than Y health, etc)

    * The inverse of a condition

    * A zero-sum condition or exclusive condition - when one player has it another loses it

    * Any combination of the above creates emergent complexity and the dimensions of analysis increase exponentially, giving players endless exploration potential

    Note: From my experience in DMing, I have to add that systemic world properties are massively immersive. Nighttime effects and weather as a system that impacts gameplay is potentially difficult but valuable to player flow.

    Gear is abstract consumables. The trigger is equipping it and the duration is until you take it off. Levels are abstract gear, in that it's functionally something that automatically equips when you level and only unequips if you delevel. If you have a health potion that gives you 60 health on a 1 minute cooldown, how is that different than a piece of gear with 1 health per second as an effect? It destroys itself, but that's the functional difference.

    Extra reading on abstraction in video games: https://www.phil-fak.uni-duesseldorf.de/fileadmin/Redaktion/Institute/Kultur_und_Medien/Medien_und_Kulturwissenschaft/Dozenten/Szentivanyi/Computerspielanalyse_aus_kulturwissenschaftlicher_Sicht/WolfAbstraction.pdf

    Efficient Set Comparision

    The second thing is you can discriminatorily truncate the set by removing items in the set that don't meet a criteria you set. If you're looking to buy a house, a house that's already nearby is a good choice and perhaps houses that are mansions are not in the budget yet. It's easy to generate a set of houses nearby by looking at realty in your city, then it's similarly easy to iterate over the set and prune based on the mansion feature. This pruning redone across many cities generates a set of sets of houses. If you take the best house out of each of those sets and generate a new set, you have a metaset of houses. This is an efficient way to delve through an any sized series of items to reach the feature optimal item in the sets. Exploring additional sets is also efficient, since you don't have to revisit all the items in the previous sets.
    7U1GDCU.png
  • Can you have a "TL:DR" for that bit too?
    This link may help you: https://ashesofcreation.wiki/
  • ImmortalmageImmortalmage Member, Alpha One, Adventurer
    daveywavey wrote: »
    Can you have a "TL:DR" for that bit too?

    It took me an hour to write, it better take you an hour to read.
    7U1GDCU.png
  • NoaaniNoaani Member, Intrepid Pack
    daveywavey wrote: »
    Can you have a "TL:DR" for that bit too?

    It took me an hour to write, it better take you an hour to read.

    As someone that is no stranger to long posts on these forums, you need to get people invested in the discussion before expecting them to read a post of that length - and even when invested,expect little more than skim reading.

    Rather than asking for a tl;dr, I am going to ask you to write a short post of maybe 3 paragraphs (aim for 200 - 350 words at most) that is there to attempt to get people interested in what it is you want to talk about. Because I am not going to invest the time reading these two posts before knowing if it is a conversation I want to be a part of.
  • CptBrownBeardCptBrownBeard Member, Braver of Worlds, Kickstarter, Alpha One
    Part 2 of 77?
  • ImmortalmageImmortalmage Member, Alpha One, Adventurer
    daveywavey wrote: »
    Can you have a "TL:DR" for that bit too?

    It took me an hour to write, it better take you an hour to read.
    Part 2 of 77?

    Both facetious comments. I thought daveywavey was joking around, so I was as well, maybe if that's kind of crazy is on the forums I can say here it's not a 77 part series.
    Noaani wrote: »
    Rather than asking for a tl;dr, I am going to ask you to write a short post of maybe 3 paragraphs (aim for 200 - 350 words at most)

    It's my thoughts on a game theory concept (information compression and filtration) through the lens of a non-video game designer just now touching on the MMO world. It's mostly likely 0% interesting to 99% of people, but for the 1% of people who would be interested in that they'd be extremely interested.
    7U1GDCU.png
  • NoaaniNoaani Member, Intrepid Pack

    It's my thoughts on a game theory concept (information compression and filtration) through the lens of a non-video game designer just now touching on the MMO world. It's mostly likely 0% interesting to 99% of people, but for the 1% of people who would be interested in that they'd be extremely interested.

    As a topic, that is something that may well have interested me, as I enjoy reading about and discussing game development theory.

    However, the way you have presented things here is of no real interest to me.

    Unless you are a very good writer, you need to bring people in to the discussion before dumping the entire discussion on them in one go.
  • maouwmaouw Member, Alpha One, Adventurer
    You have info, but it's not clear what the driving point of your discussion is (usually a discussion is spawned from a question of some sort and seeks an answer).

    Some suggestions:
    • At what point does data abstraction go too far?
    • I think data abstraction is really useful for pruning ideas down to their core identity (and therefore better iconography). Do you agree?
    • Data abstraction is useful for creating non-linear variance in the game, which makes the game more fun. What parts of AoC are too linear?
    I wish I were deep and tragic
  • ImmortalmageImmortalmage Member, Alpha One, Adventurer
    daveywavey wrote: »
    You're supposed to put a "TL:DR" in there.....

    Efficient Set Comparision

    The second thing is you can discriminatorily truncate the set by removing items in the set that don't meet a criteria you envision. If you're looking to buy a house, a house that's already nearby is a good choice and perhaps houses that are mansions are not in the budget yet. It's easy to generate a set of houses nearby by looking at realty in your city, then it's similarly easy to iterate over the set and prune based on the mansion feature. This pruning redone across many cities generates a set of sets of houses. If you take the best house out of each of those sets and generate a new set, you have a metaset of houses. This is an efficient way to delve through an any sized series of items to reach the feature optimal item in the sets. Exploring additional sets is also efficient, since you don't have to revisit all the items in the previous sets.

    Expanding on this repeated filtration, this type of thought extends past pulling real results from real sets. This is a powerful method of abstraction that maintains coherence throughout each iteration of the process. The issue with most abstractions is it loses precision, and the problem with direct interpretations you lose the values that exist only in other dimensions of analysis.

    To get a bit meta, that's the difficulty in providing examples for these thoughts here. An idea can be representative of a large swath of things, but once you provide an example you hone in on the one concrete thing that is real in the set of things the idea could be and you lose out on all the other things it could be.

    One tool I've used as a DM is Venn Diagrams, and often focusing on the symantic intersection (https://en.wikipedia.org/wiki/Venn_diagram#/media/File:Venn0001.svg) and symmetric difference (https://en.wikipedia.org/wiki/Venn_diagram#/media/File:Venn0110.svg). In many designs, you want to know how things overlap to get a better understanding, and allow for extensions of the space of things you're trying to create more of. The second most common thing for myself is looking at symmetric differences in order to exaggerate the differences of things in a set. This might look unwieldy when going up to 3 items in the set that's being compared (https://en.wikipedia.org/wiki/Venn_diagram#/media/File:3-set_Venn_diagram.svg), and is definitely unwieldy when approaching more than 3 dimensions (https://en.wikipedia.org/wiki/Venn_diagram#/media/File:Venn_0001_0110_0110_1000.png). A great trick to explore hyperspace is to flatten it out and just extract the operation being performed, which for Venn Diagrams is the inclusive match or the remainder (which functions as the symantic intersection and symmetric difference respectively). It becomes so simple to solve, solving it looks like a list again. You can do this to create something that is an abstraction of the things in the set. This is an oversimplified characteristic map of a class derived from a snake, eagle, and fire, and although the aesthetics may be easy to imagine with a fiery class, this is the simplistic list where I just brought over each trait in a row that matched along each column of the features:

    TL;DR You can take things and mash them together or pull them apart to create really cool new things. For example, a dragon is basically a man and threats. A dragon is greedy; he sleeps on piles of gold he can never spend. He's gluttonous; he eats entire fields of livestock depriving the villages. He's slothful; sleeping for centuries at a time. He's wrathful; killing rampantly and burning rampantly. He's prideful, lustful for dominance, and is a body of threats. Threats like snakes, fire, and predators of the sky. In all things there is power in abstractions and oftentimes the abstractions are better than the dullness of precise reality.
    7U1GDCU.png
  • RavudhaRavudha Member
    edited December 2020
    .
Sign In or Register to comment.