Talk:Procedural generation

Older comments
This needs to be changed. It is Spore specific right now and copys directly from the Spore entry making it redundant since thats probably the only place that links here.
 * It's a very interesting concept that will become more and more games will use. I'm removing the cleanup tag. Jacoplane 19:19, 31 July 2005 (UTC)

The article may be brief, but its spore specificness is only due to a lack of examples. It deserves an article of its own, however, as it is an entirely different concept to procedural textures.
 * There's also the article on Procedural programming.

Darkling Simulations DarkTree is another popular middleware for procedural texture/shader synthesis, worth a mention if SpeedTree is featured. Savuporo 14:23, 12 January 2007 (UTC)

Personally, I don't see the relationship between procedural generation/texturing and procedural programming. Sure, you can use a "procedural-programming language" to do procedural texturing and such, but it's just like saying you can write programs using C/C++. By the way, that article is not really useful in my opinion.

Anyway, the point is that the relationship is too weak for me. It's like saying that, if you don't understand the working of C/C++ you should check out imperative languages. Sure it helps but the concept is on two totally different dimensions. It's more likely the average user will end up messing various concepts.

The two are getting linked just because they share the word "procedural" but as I see it, the meanings of that word is different in the two contexts. This is very unfortunate since I hardly believe any useful procedural generation algo (SpeedTree comes to mind) can be written using procedural programming methods. MaxDZ8 talk 11:29, 6 February 2006 (UTC)
 * I think you are correct. The procedural programming page should not like to procedural generation in the See Also section. The two are unrelated in any meaningful way. Jessmartin 06:00, 17 February 2006 (UTC)

Revert pending for User:Ashley Y version. The user linked this page with functional programming. I am going to revert it for reasons similar to above. This time, it's claimed procedural generation to be a subset of functional programming. I consider the linkage to be way too weak. If procedural programming has been removed, by sure means this also should be. To re-state a concept above: no doubt you can do procedural generation with this paradigm but it's way too general. There are also a lot of real-world scenarios in which the whole thing just does not fit. Last but not least, the reading of the linked resource does not seem to be very useful and the category added is about heavy mathematical wizardry. There's nothing wrong in that but this makes me think this would be out of place. 83.176.85.211 08:58, 13 March 2006 (UTC)

Merge comments
I believe this shall survive the merge with procedural synthesis. Although as far as I know "synthesis" is the correct term, it's more likely average Joe will like "generation" much more. I'll work on a offline merge in the next few days and propose the results soon. MaxDZ8 20:52, 28 November 2005 (UTC)

I've merged the two pages. I believe this shall now be merged with procedural texture, mainly because the other is very short and I believe it's good to have few big articles than jump here and there. Texturing is just an application so it's also another good reason to merge it. MaxDZ8 18:43, 29 December 2005 (UTC)

I've noticed the proposed merge with procedural texture has been removed. I'll take this as proof some people wants to keep the two pages separated. I believe I'll spend some time on the other page in the next few weeks. MaxDZ8 09:04, 13 January 2006 (UTC)

Other Procedural Generation methods and related topics
There are some other procedural methods which should/could be mentioned in this article:
 * L-System - Procedural method for generating trees. This should particularly be mentioned in the speculative paragraph about trees.
 * Fractal landscape - Procedural method using fractals for generating terrain.

Quite a few other things relate to these methods as well. Procedural textures and Perlin noise are just a few examples. This article could ideally be restructured to include a more general overview of the variety of techniques used in this area as well as some applications.Jessmartin 06:08, 17 February 2006 (UTC)

I don't have much knowledge of LSystems and I only know basic facts about fractal landscape generation but the actual definition of this latter thing seems to be quite different from what I know. Linking to procedural texturing has been considered as you can see above, but then I moved on editing another page with higher priority (this is "quite nice" actually) and forgot completely to blend it in the text. Same fate happened to Perlin noise references.

It takes a little bit of extra effort than adding a few links or embedding the references in the text nicely. I discussed on the help desk some time ago the need to supply code pieces to demostrate some "almost real" examples but we couldn't find an useful agreement on the language or technologies to be used. This annoyed me pretty much and made me focus on another article. Now that you're providing extra knowledge, I am confident we could finally raise the bar on this topic to useful levels.

There's also the problem that I kept a lot of content from previous revisions (see the blue box?) but I wasn't able to verify it and I personally dislike the whole chapter. What do you think about this issue?  MaxDZ8 talk 22:27, 3 March 2006 (UTC)

Example of procedural generation in DOOM
In DOOM (the 1993 version), iD software had a feature called "auto-pitch" I believe, which adjusted the pitch of sounds randomly when the sound was played. Forexample so that gunshots would sound different, or when selecting stuff in the menu.

I'm not sure I remember this one but I'd rather look at it as a sound processing effect and as far as I remember the sounds were taken from pre-made files so I don't see the connection with procedural generation. Thank you anyway for pointing this out here.  MaxDZ8 talk 16:14, 6 March 2006 (UTC)

This thing...
Procedural generation as an application of functional programming

Procedurally generated content such as textures and landscapes may exhibit variation, but the generation of a particular item or landscape must be identical from frame to frame. Accordingly, the functions used must be referentially transparent, always returning the same result for the same point, so that they may be called in any order and their results freely cached as necessary. This is similar to lazy evaluation in functional programming languages. I understand this is page actually unverified but I still think something should be put here with salt. Regarding the above statements, I would say that: Could you please point out the goal of that?  MaxDZ8 talk 10:49, 21 March 2006 (UTC)
 * This is way too short to be a chapter by itself.
 * The statements introduce external concepts (functional programming, lazy evaluation, referentially transparent), this would be good but I personally believe those are way too far from the topic.
 * Some of those properties are adviseable but not always required.


 * Procedural generation happens to be an application of functional programming, and really ought to be called "functional generation". The problem is that the people who write procedural generation code don't seem to be very familiar with functional programming, so the connection isn't obvious to them. They may create generation functions and be aware that the functions will usually have to have a certain kind of property, but may not be aware that computer scientists refer to that property as "referential transparency", and so on. &mdash;Ashley Y 09:17, 10 July 2006 (UTC)

Your point is taken. I admit that having being self-formed, I lack the academic ... things (I have however used functional languages for almost three years). Yet, I still think this degree of abstraction to be far from correct (as I disagree with most academic stuff I know about). Why? Because there's nothing like "functions" inside a computer. It happens that a series of binary tokens works so similar to numbers and operations some math guru thought it would be nice to estabilish a link between math and physical computers. As now, I think the model envisioned to be actually more involved by the facts themself, thus defeating the need for a model.

After having read your rationale, I am convinced you know the issue. I won't revert anymore your additions, but I still disagree with this. Do as you feel better, however I still think that working with abstraction levels, hypotesis and audience, in computer science is possible to say everithing and the opposite of it, yet hit the truth.

 MaxDZ8 talk  08:13, 11 July 2006 (UTC)

This assertion is simply not true. There is NO requirement that all procedural content be containable in an explicit function or re-evaluatable for the same result. That's because not all assets are generated each frame. Certainly in real-time applications, one does not perform the generation on a per-frame basis. Furthermore, the reference to functional programming is far less relevant than a lot of the basics that are already being left out here. I have *never* heard anyone in the field make this comparison (and I'm familiar with a lot of the literature). Can you cite references? --XeonXT (talk) 07:11, 26 October 2011 (UTC)

Landscape/Terrain Generation
Would an article detailing terrain generation fit as a stand alone article on Wikipedia? Since there's a lot of terms for it (pretty much any combination of "procedural"/"generation"/"synthesis"/"terrain"/"landscape"), what would be an appropriate title? Indeed, there is already an article on procedural textures, although it does seriously need more content. In any case, there are already several articles concerning terrain generation algorithms and a "parent" article is sorely needed. I'm pretty sure I'd be able to fit in a lot of content. Nezbie 05:10, 2 May 2006 (UTC)

I'm not aware of the other articles... could you point me to them? To reply you I would like to take a look.  MaxDZ8 talk 15:53, 2 May 2006 (UTC)


 * I wonder if there are any orphaned articles on the subject. In any case, some relevant articles include: algorithms (Fractal landscape, Diamond-square algorithm) and software (MojoWorld, Terragen). The applications and widespread use of terrain generation alone should make the topic worthy of a full-fledged article, including elaboration of some of the most used techniques (including subdivision algorithms, frequency synthesis, faulting, noise, etc...). Nezbie 18:39, 2 May 2006 (UTC)

I agree that this topic really needs its own page. Covering this here is definetly brain-damaged, considering the amount of detail needed. There will probably be some merge/redirect to do... fractal landscapes are procedurally generated landscapes for example. Maybe the topic needs even more than a page (algorithms and software perhaps).

It looks a dauting task. Good luck!

 MaxDZ8 talk  17:33, 3 May 2006 (UTC)

Sound
This sentence didn't make sense to me, so I removed it: "it has been employed in many songs (for example The Cure's songs largely employ synthesizers while techno music is often largely synthesized)". Creating electronic sounds seems a far stretch from procedural generation. (Feel free to correct me, of course.) Tlogmer ( talk / contributions ) 16:22, 6 May 2006 (UTC)

To tell the truth, I has been told about this by a friend and I really cannot prove it so I agree with this.

As for the citation needed, what actually needs clarifications?  MaxDZ8 talk  10:37, 7 May 2006 (UTC)
 * 1) Synthesis VS generation?
 * 2) Electricity + creativity = Synthetized sound?

I am not an expert in procedural generation/synthesis; to my understand is that procedural systhesis, in computer graphics, is when you use some sort of algorithm to generate graphical content in real time rather than taking a "sample" of texture or use a pre-computed information. If my understanding is correct; I would like to make a clarification as how the procedural synthesis in graphics can be compare to sounds. without goin' into digital vs. anglog sound synthesis details, typically, there are two common way to create sounds: synthesis and sampling. synthesis is when you "generate" a waveform via oscillator; then pass it through bunch of filter to produce a desired sound (waveform). you can say that the procedure is done in real time as you can modify your originating waveform in your oscillator directly; the changes made is immediately effects your output sound. sound producer use various tools to shape their output sound such as different originating waveform (sin, sqaure, saw) or different filter (low pass, cut-off). thus, by saying that electronics sound is created by "electricity" and "creativity" is not entirely wrong. on the contrary, another way to create sound is "sampling" or "sample playback". this is done by first "sampling" the waveform from source audio, store it, and use osccilator to change the frequency (pitch) during the playback. this is similar to scaning a texture from real material and use it in 3D engine. if the analogy is texture reconstruction VS. procedural synthesis; then to sound it is, sound synthesis VS. sample playback. hope it helps. underexpose 01:51, 13 July 2006 (UTC)

Isn't this also used extensively in "Rogue-like" dungeon-crawling games?
AFAIK many rogue-like games (like Nethack or Moria) generate their dungeons "randomly"... Does this also fit into this category, and an someone confirm this? 193.96.240.2 08:24, 15 June 2006 (UTC)


 * I also think The Elder Scrolls II: Daggerfall may have used this process. The majority of its cities, landscape, and dungeons were generated at the start of a new game rather than beforehand. Zenithan 19:35, 25 August 2006 (UTC)

The "Console Support and Speculations" section seems totally out of place, procedural programming has nothing to do with the hardware specifications of all those console cpu's. 20:40, 17 January 07 (GMT)

I agree! Removed.

 MaxDZ8 talk  15:42, 18 January 2007 (UTC)

dodgy link
[Free access link found!]

about the first external link:
 * Real-time procedural generation of `pseudo infinite' cities - ACM

While this seems to be a very important paper on the topic, it is not available freely. I have a free web account with ACM and this text cannot be accessed with it, leaving only the possibility of purchasing it for $10.

Thus, there are some problems with this link, probably violating wp:web:
 * 1) It could be considered spam/ad, since there's no more info on the topic behind the link without having to register&pay for it.
 * 2) It isn't easily accessible, rendering it useless.
 * 3) There are no hints toward the delicate nature of the link accompanying the reference.

In summary, this link should be removed, at best substituted with a better link to the paper, which is probably hard to find.

Opinions? --Schmiddtchen 12:03, 13 February 2007 (UTC)


 * I'd have to agree, that link is pretty useless. Virek 13:05, 2 March 2007 (UTC)

I managed to find a link to what should be the same: http://www.infinitylab.com.au/research/graphite2003.pdf 194.237.188.194 (talk) 07:14, 26 May 2008 (UTC) (Did i get everything right when making this 'post' so to speak?)

Real world ref to wiki
This month's Game Developer Magazine references to this article in its "Inner product" section! Congratulation to the editors! This should be taken as an indicator this is actually "good"!

85.18.201.168 11:41, 8 March 2007 (UTC)

Generation v. Synthesis
If "procedural synthesis is actually the correct term" then why is the entry still called procedural generation. Shouldn't the entry be under the "correct" title of procedural synthesis, not a synonymous title and having the "correct" term redirect here. Just a thought, correct me if I'm misunderstanding something.
 * 1) I have found no proof of the actual "superiority" of a term over the another (besides my knowledge).
 * Well, if there is no actual superiority between the terms, then why do you clearly assert the "correct"ness of synthesis in the second paragraph? It either is or is not more "correct", and if it is more "correct" then my feeling is that, regardless of publicity, the article should be titled by the "correct" term. In my opinion, one of two things should be done:
 * Move the article to Procedural Synthesis and state in the introduction that "although procedural generation is the more popular term, procedural synthesis is actually the correct term." or...
 * Remove the statement stating the correctness, and therefore superiority, of synthesis.

 MaxDZ8 talk  08:58, 9 March 2007 (UTC)
 * New: considering this has been reported to the press (see above) it doesn't seem rationale to change it now.

Ok, removed. I wonder however why you couldn't to it in the first place.

 MaxDZ8 talk  08:28, 24 March 2007 (UTC)
 * Common courtesy, and I wanted someone more knowledgeable to make the final choice. MadScientistVX 23:13, 24 March 2007 (UTC)

I see. Thank you... but I don't really consider myself an expert at this.

 MaxDZ8 talk  10:18, 25 March 2007 (UTC)

Fractals
"This concept takes on dramatic expression in fractals, a specific example of procedural generation; one around which a whole body of mathematics&mdash;fractal geometry&mdash;has evolved."

Forgive me if I am mistaken, but I was under the impression that fractals in general are not defined as "procedurally generated", though for the purposes of use in computer art or games they may be. Furthermore, I am certain that the mathematical study of fractals (the fractal geometry link is redundant, by the way) did not arise as a result of the development of procedural generation. Yes, fractals as we encounter them in computer graphics are generated by a procedure or algorithm, but that does not make them procedurally generated per se. I hope you see my point (or even read it, for that matter). Robin S 01:28, 19 April 2007 (UTC)

Well, I'm not sure I understand. You say that it comes from an algo so it is not procedurally generated because it's more than just procedurally generated? I believe it's a special case.

 MaxDZ8 talk  05:51, 19 April 2007 (UTC)


 * I'm saying that something which can be produced by an algorithm is not necessarily "procedurally generated". Looking at the definition given in the first line of the article, procedural generation is used to create content on the fly, whereas fixed and well-studied mathematical objects such as the Mandelbrot set will be exactly the same every time they are generated regardless of the details of the algorithm used to generate them and thus should not, in my opinion, qualify as content "created on the fly". Moreover, the mathematical study of fractals (as referenced in the quote, above, from the article) is independent of the algorithms used to generate them. The two are separate fields. Fractals have applications in procedural generation but this does not make them a subfield of the latter.


 * edit: if you wanted to be pedantic, you could say that virtually every computer application ever written uses the principle of procedural generation to an extent, because much of the content is generated at runtime (for example, the exact appearance of a window will be altered as it is resized). I feel, however, that this is missing the point. Robin S 20:08, 19 April 2007 (UTC)

I believe I got it but frankly, I don't know enough!

 MaxDZ8 talk  07:10, 20 April 2007 (UTC)


 * The Houdini/Prisms 3D animation software is a 20 year old procedural 3D animation program that has been used in dozens of films and commercials. http://en.wikipedia.org/wiki/Houdini_(software) It includes tool sets for procedural modelling, animation, shading, rendering, lighting: http://en.wikipedia.org/wiki/Houdini_(software) —Preceding unsigned comment added by 24.21.139.14 (talk) 04:17, 7 September 2007 (UTC)

Organization of Software using procedural generation
I hope nobody is upset, but I feel that the section looks better the way I organized it. PhycoFalcon 05:33, 24 September 2007 (UTC)


 * I found a reference for Jade Cocoon. Unfortunately it's on GameFAQs (http://www.gamefaqs.com/console/psx/file/197666/10553), so it can't be hotlinked Eoseth (talk) 07:58, 30 April 2008 (UTC)

Image copyright problem with Image:Kkrieger screenshot 01.jpg
The image Image:Kkrieger screenshot 01.jpg is used in this article under a claim of fair use, but it does not have an adequate explanation for why it meets the requirements for such images when used here. In particular, for each page the image is used on, it must have an explanation linking to that page which explains why it needs to be used on that page. Please check


 * That there is a non-free use rationale on the image's description page for the use in this article.
 * That this article is linked to from the image description page.

This is an automated notice by FairuseBot. For assistance on the image use policy, see Media copyright questions. --04:09, 18 September 2008 (UTC)

Runescape
I don't know how exactly it works, but Runescape uses a procedural graphics rendering technique which produces (as far as I know) all models in the game. As it's closed source it's difficult to know the extent, but it clearly shows that the technique is usable.

Despite WP:BOLD I didn't want to add it, because I'm not particularly sure where it goes. If a graphics guru is amongst the people here, then here's an example Philipwhiuk (talk) 20:23, 18 August 2009 (UTC)

Star Trek Online
Star Trek Online has a similar "Genesis System", that generates stars, planets and species, but keeps creating these worlds for endless exploration (Similar to Spore). If a player wants to keep a system they find, all they do is they save the coordinates so other players can find them. So would Star Trek Online be included in this list, or is the Genesis System different to Procedual generation (Im not very technical when it comes to programing)? 88.105.124.221 (talk) 20:40, 23 November 2009 (UTC)

Notable Examples
As far as procedural content for video games go, Castle Wolfenstein predates every game listed with the possible exception of roguelikes (Beneath Apple Manor and Rogue specifically). Some other early examples: Gateway to Apshai which procedurally generated dungeons based on a fixed random seed table (interestingly one of the seeds causes the Atari and C64 versions of the game to crash); The Seven Cities of Gold (video game) which generated an entire New World to explore for Spain. Xot (talk) 12:19, 2 December 2009 (UTC)

Serious Editing Needed
"Procedural generation is a widely used term in the production of media, indicating the possibility to create content on the fly rather than prior to distribution."

This is wrong. It does not matter whether procedural content is generated before or after distribution. Procedural content is produced algorithmically, rather than manually. Of course, every way handling data on a computer requires some sort of algorithm, but that is the best division that can be made. If anyone chooses to generate procedural content, and distributes it after some editing, it may still be largely procedural content.

Also the article in its present form focuses far too much on computer games, which are just one application. I propose to delete the article and link to procedural modeling instead.

87.153.48.240 (talk) 17:04, 6 November 2009 (UTC)


 * Not wrong. Slightly misleading, but technically accurate.  Read carefully, with added emphasis: "indicating the possibility".  With procedural generation, it is *possible* to create content after distribution instead of before.  You don't have to, but you can.--99.17.1.168 (talk) 04:43, 7 January 2010 (UTC)

External Links Fixed
I replaced a few of the broken links with the same content from alternate sources. For the link to the book "Texturing & Modeling - A Procedural Approach" I have put a link to a website of one of the authors that talks about an older version of the book. It's far from perfect, but at least this link works. I also added in one link to an article from Gamasutra that I thought was informative FatPope (talk) 02:21, 6 March 2010 (UTC)


 * Also added in a link to 'FUEL', with appropriate reference --FatPope (talk) 02:51, 6 March 2010 (UTC)

Cellular automata
Cellular automata, along with fractals, are good examples of procedural generation. However, this section reads more like an "advertisement" of the topic, meant to convince game designers to use CA. Are there any actual examples of CA being used in games or for other kinds of content generation? EverGreg (talk) 16:27, 21 July 2010 (UTC)

Agreed. I've only ever seen CAs used in a dynamic cloud system before, but they're not at all commonplace in procedural generation. "Fractal" terrains and perlin noise textures are far, far more common...so it makes no sense to go into detail on CAs when the most basic methods have not been explained in detail. --XeonXT (talk) 07:17, 26 October 2011 (UTC)

Film subsection
The "film" subsection is completely uncited and, quite frankly, unintelligible. Although someone else has commented elsewhere on this discussion page that the Houdini software has procedural generation tools, I'm pretty sure that's more for like adding dynamic "noise" or "grain" to an image. The whole "imperfect factory" thing is nonsense and unless anyone can clean this up I'm going to remove it. Alphachimera (talk) 14:39, 5 July 2011 (UTC)
 * I think the film section is pretty bad. Peter Jackson's King Kong had a procedurally created 1930s New York City. There are certainly many more examples, and the "imperfect factory" needs to be sourced. I've never heard of this.


 * Houdini's proceduralism goes way beyond image manipulation. Almost everything in the software is accessible as a node and can be created or modified procedurally. You could for instance use procedural patterns to make a black and white image. You could then use nodes to convert the image into curves. Use a procedural modeling network to turn the curves into 3D roads and buildings. Use more networks to apply materials and shaders etc. Chris TC01 (talk) 17:05, 5 July 2011 (UTC)


 * I had no idea. I just found some pretty interesting youtube videos of how it can be used to create buildings, I'm guessing it's not that much harder to create a whole city. However, I'm starting to wonder if there's an important distinction between procedural _generation_ to create from-scratch elements and procedural _modeling_ to fill defined spaces with particular, constrained elements. Maybe it's my POV and my limited experience but creating a realistic 1930s New York seems sooooo different from, say, populating new areas in Minecraft or new solar systems in Star Trek. Alphachimera (talk) 19:34, 6 July 2011 (UTC)

Minecraft Isn't a Strategy Game
It should be listed under Sandbox or something. — Preceding unsigned comment added by 81.228.145.220 (talk) 02:28, 26 August 2011 (UTC)

Coded arms random levels
I'm 99% sure Coded Arms and its sequel Coded_Arms:_Contagion have random levels. — Preceding unsigned comment added by 76.121.154.215 (talk) 08:24, 29 September 2011 (UTC)

Trim down the list
One of the issues with this article seems to be the length of that list. It actually links to a category, so why does it need to list so many games that make use of procedural generation? Does it aid in understanding what procedural generation is about? It also results in a lot of odd categories that don't always make sense.

I suggest using the games in this list to fill out the category some more. Otherwise, may I trim down that list? I don't want to delete half of the list without hearing any opinions. Maplestrip (talk) 09:24, 3 November 2014 (UTC)

After reading WP:BB, I decided to just go for it. I've only removed unreferenced material, and not nearly all of it. Still managed to remove at least a third of the list. Should I also completely remove the shooters and strategy games? Has anyone even tried to look up references for those? The arcade games seem important to keep in for historical reasons, but those too really need references. Maplestrip (talk) 12:50, 4 November 2014 (UTC)

I added a note that procedural generation in games originated in D&D.`
I am a new Wikipedia user so if I have done something horribly wrong, please let me know. :)Charles999JF (talk) 16:03, 7 October 2019 (UTC)
 * You need a source for that. The random encounter/weather/etc. tables in D&D are just that - random, they don't represent procedural generation. --M asem (t) 16:31, 7 October 2019 (UTC)

I cited 2 books. I have used these tables myself and they are absolutely procedural by the existing definition you put in this article.Charles999JF (talk) 16:40, 7 October 2019 (UTC)
 * You can't directly site the books that are being used as examples as a source to say they are procedural generation - that is original research. However I was able to find at least two academic papers that credit D&D and other tabletop systems for procedural generation, so I've added those and sourced appropriately. --M asem (t) 17:03, 7 October 2019 (UTC)

Categorization in applications of randomness
Hi Hellknowz,

I'm not an expert on categorization, but sparse categorization on articles seems less useful for readers (at the very least personally speaking). Currently, this article is only categorized in Category:Procedural generation.

Also, in your revert you said it is already part of Category:Procedural generation subcat, but it's only part of that category because I added it today.

I was mulling over whether the procedural generation category should be CfD nominated, but regardless I feel it's useful for this article to be categorized in Category:Applications of randomness

-- Retro ( talk  &#124;  contribs ) 13:37, 19 July 2022 (UTC)


 * Additional note: it appears there are some specific guidelines for eponymous categories: WP:EPON. It provides three ways for handling the situation, and for our purposes, I think #1 makes sense. A strictly nonredundant category hierarchy isn't an absolute rule. Retro  ( talk  &#124;  contribs ) 13:52, 19 July 2022 (UTC)


 * Hi. You're right, I didn't realize this. I've never actually encountered this sort of case. The part I was referring to is the WP:SUBCAT, i.e. "pages for sub-categories should be categorised under the most specific parent categories". But here I agree that WP:CATMAIN applies, i.e. "articles with an eponymous category may be categorized in the broader categories". I reverted my undo. I don't have any comment on sparse categorization - that's not really a factor for anything as far as I know. Some topics just don't have categories. — HELL KNOWZ ∣ TALK 18:28, 19 July 2022 (UTC)