Drop me a line or reply to this post if you want access to the Google Analytics information for pcg.wikidot.com. Unfortunately, the permissions model for Analytics as I understand it means I can only give View Only access without having to throw away all the historical information and starting again.

Regards,

Andrew Doull

Julian Togelius has set up a Google group for PCG at http://groups.google.com/group/proceduralcontent which appears to already have more discussion other than housekeeping than the forums here. I'd like to invite regular contributors to join the group - I suspect we should perhaps have a brief poll here on whether we should close down the forums here other than the General Discussion forum.

Regards,

Andrew

I'd surely consider AI content though, the difficulty and complexity of which certainly affects the enjoyability of the game. I'd say only include games that feature adaptive AI as a major feature, like Left4Dead.

On the topic of Left 4 Dead, could you point me to a description of its adaptive difficulty, if you happen to have one? Because my limited experience with the game seems to suggest that what the AI Director is more an example of Algorithmic Difficulty, rather than adaptive….if I try to play a campaign on a difficulty that is too high for me, I will just fail the campaign. Likewise if I play on a difficulty that is too easy, I will run through it unchallenged. The AI Director may place the zombies in new locations each time, or order in zombie waves in different locations, but if I opt to play a hard campaign, then the AI Director will algorithmically give me a hard campaign.

A lot of the press about Left4Dead mentioned it's adaptive difficulty - [http://www.edge-online.com/blogs/gabe-newell-writes-edge] is probably the easiest example to find. I'll add it to the article.

From the 'What PCG is' page, you have created this as your base definition of PCG:
"Procedural content generation (PCG) is the programmatic generation of game content using a random or pseudo-random process that results in an unpredictable range of possible game play spaces."

A potential problem with adaptive difficulty as a PCG technique is that rather than allowing the player to explore an 'unpredictable range of possible game play spaces,' the goal of adaptive difficulty (as I understand it) is to constrain the player into the game designer's preferred play space. That is, if you are too good at the game, the game will get harder, so that you progress as expected. The same is true of the 'too bad at the game' case.

Take Resident Evil 4, for instance. Much of the tension, the overall 'experience,' in the game comes from managing a small amount of resources, primarily ammunition. Therefore, if I, as the game designer, want to ensure that a certain scene or boss 'feels' right, I need to make sure that the player has a given amount of ammunition when entering the fight, so that they get the experience I designed for them. By adding adaptive difficulty, I introduce a nice negative-feedback loop which prevents players from straying too far from this 'optimal' play space, by either removing excess resources if the player is too good (and thus has an abundance of ammo) or by adding in resources if the player is too bad (and is running the risk of leaving the game out of frustration).

Next look at Oblivion. The adaptive difficulty in this game, as far as I understand it, scales the monsters' levels with that of the player. So if I approach a tower at level 1, I will fight a level appropriate enemy. Similarly, if I approach that same tower twenty levels later, the enemies will have leveled up appropriately (perhaps the bandits have full plate mail and magical swords instead of cheap daggers and torn clothing). Now, again, this adaptive difficulty serves to focus the game into a very comfortable and safe player space…I will always be able to vanquish my enemies, provided I put in approximately the same amount of effort as I did previously. This adaptive difficulty removes the player's ability to explore the more 'dangerous' parts of the play space in favour of keeping the game friendly and playable. Now, this may have benefits for many players (I personally disliked the leveling world), but as far as resulting in the exploration of an "unpredictable range of possible game play spaces,' Adaptive Difficulty, as I have seen it implemented, seems to be designed to do the exact opposite: to allow the designer to better control the player experience.

On the topic of Left 4 Dead, could you point me to a description of its adaptive difficulty, if you happen to have one? Because my limited experience with the game seems to suggest that what the AI Director is more an example of Algorithmic Difficulty, rather than adaptive….if I try to play a campaign on a difficulty that is too high for me, I will just fail the campaign. Likewise if I play on a difficulty that is too easy, I will run through it unchallenged. The AI Director may place the zombies in new locations each time, or order in zombie waves in different locations, but if I opt to play a hard campaign, then the AI Director will algorithmically give me a hard campaign.

Anyway, I hope I've at least provided a decent springboard off of which to continue thinking about the issue of adaptive difficulty as it relates to this wiki.

I've hit a stumbling point writing up an article on adaptive difficulty - always a controversial point in games. I'll quote the whole article to saving you having to go to the original link:

Adaptive difficulty is the process of adjusting the game in reaction to the player. By spawning new enemies or powering up existing enemies if the player is progressing quickly through the game, or by decreasing the frequency and/or difficulty of existing enemies if the player appears to be having problems progressing, adaptive difficulty techniques attempt to create the 'optimal' game experience.

Classically, adaptive difficulty has been seen as a hard problem, requiring a level of artificial intelligence in the game to attempt to model the player to attempt to determine if they are finding the game easy or difficult.

However simpler RPG style mechanisms can also be seen as adaptive difficulty techniques. Allowing the player to level up by playing through additional easier content can ensure the player is able to grind their way through parts of the game in order to decrease the difficulty of sections of the game where the difficulty level increases. Paradoxically, adaptive difficulty techniques which increase the difficulty of the game by scaling up enemy strength have been fiercely resisted by RPG players, as can be seen by the negative reactions to the difficulty scaling in Oblivion.

Adaptive difficulty is not usually seen as a procedural content generation technique, but it has most of the features of such techniques. It could be seen as decreasing a game's randomness instead of increasing it which would make games which feature it without other PCG features to fall outside the 'canon' of PCG games.

Should I include games which have adaptive difficulty in the PCG wiki? There are plenty of examples of games which have adaptive difficulty and are procedural (Oblivion I've already mentioned, Left4Dead) but there are plenty of games which are not (SiN: Episodes). And I don't want to include every RPG, based on the argument I've made above.

The real question is not whether I should include these games, and the answer to that is probably not, but why? What good reason can I give to not include SiN: Episodes, for instance, as a procedural game?

Again, the randomness argument is the most plausible, but it is not completely convincing.

Convince me.

I'd love to be there…

It really happens though. In X-Com a soldier had the same name as a friend I knew, so he was kept around, even after I discovered PSI and found that he had mediocre PSI strength.

Many of the TIG-PCG entries got attributes for items by asking the player for a word and seeding the RNG with it, or hashing it, etc.

Diablo adds prefixes and suffixes to an item, and the mods on an item are dependent on the prefixes/suffixes.

Then there is Treasure of the Rudras where the player-entered name of the spell was parsed into prefixes, base, and postfixes to select the effect of the spell.

I'll add in the additional reference I was planning to create on a name generation page to this one…

Thanks to whomever set those up.

Having set up a fairly rigid taxonomy of thinking about PCG types, I think it's time to break it down. There's a couple of reasons why:

1. It's too difficult to spell taxonomy.
2. I've created this pcg-algorithm: _concepts tag and want to use it.

So if no one has any objections in the next couple of days, I'll move these across.

Of course, there's an equally valid argument to go 'why are these algorithms at all?'. They're not. They're higher level concepts. But they lie close to algorithms on the thinking about 'how do I implement the creation of procedural concept.'

Thoughts?

Andrew

The FlickrGallery module is kind of neat but could still use some work. There are two features that should exist, might not take to much work on the part of wikidot:

1: allow tags to be negated: that is we want to exclude all images that have certain tags. This isn't a big issue but would be nice. I filed this as a bug but that probably isn't the correct place to put it.
something like:

 [[module FlickrGallery tags="procedural" excludeTags="unicorns" tagMode="any"]] 

2: allow tags to be combined with "alternative attributes": photosetId, groupId, groupUrl. If this was implemented we could use the gallery to populate other pages. (An expamle is fractal. Using tags only there are many images there that are not proceduraly generated (such as photographs of plants) and very few have any explanation as to where they come from. There are images in the flickr group pool that have the tag fractal and it would be nice if we could show them that page.)

 [[module FlickrGallery groupId="1153986@N23" tags="fractal"]] 

Whether a wiki should in and of itself be the vehicle to promote discussion, I am not so certain. Perhaps more proactive endeavours better suited to engaging people would be better suited. One thing that comes to mind is perhaps informal challenges where people are encouraged to implement some usage of a procedural generation technique, in friendly competition.

Anyway, this is my plan. Any thoughts?

At this stage I don't know if we have enough content yet for a clear style to have emerged… but a less formal style is fine by me.

I am not the first to speak of this:

Flammifer:

I gotta admit, the Wiki as it is doesn't really encourage community discussion. It has a bit of an encyclopedic tone, it might work better with a more informal tone, like tv tropes and, well, most good non-wikipedia wikis :)

andrewdoull:

Feel free to give the place more of a community vibe.

So we should be using an informal style. Don't take the long, dry, and boring way to explain something just because this is a wiki. Say it as it is, or at least say it how you would normaly. No need to use fake voice. This is the tone. I wish I could be more specific, but this is a bit unclear to me.

On the other hand, I want my writing to be written for the web. This is the style:

• Consise text. Fewer words are better words. Make every word count. (Strunk and White FTW!) Use half the word count (or less) of conventional writing. This could compete with the tone but if it makes the content easier to understand then all is well.
• Scannable layout. Use bulleted lists where they work, bold key phrases, and otherwise make it easy to find out what an article is about without actually reading it. This means
  • highlighted keywords (hypertext links serve as one form of highlighting; typeface variations and color are others)
  • meaningful sub-headings (not "clever" ones)
  • bulleted lists
  • one idea per paragraph (users will skip over any additional ideas if they are not caught by the first few words in the paragraph)
  • the inverted pyramid style, starting with the conclusion
• Objective language. Prefer neutral, descriptive language over subjective boasting. Marketese (aka doublespeak) causes mental friction, it slows down reading and sows distrust. Again this could compete with the tone but it shouldn't. The big point is to be honest and don't sound like you are selling something.

Anyway, this is my plan. Any thoughts?

Maybe link to one or two of the You Tube videos that they have, and make that a news item instead?

a) the sounds that make up the music are not procedurally generated
b) the music selection is procedural, made up of a number of music fragments put together.

In particular, key words: dynamic and artificial life strongly suggest that procedural algorithms are at play.

I also think this long form style comment deserves to be an article: we should really have a section of is 'x' procedural articles, because these are thought provoking and good examples of stuff you may not have initially thought was procedural.

Structure is a procedural system generating virtual cities including destroyable building interiors. For rendering of the massive city the system is using OpenGL and ARB_occlusion_query.

The actual site for Structure is missing its home page, so I can't add it as a news item.

We can't look at the source code to check, we must rely on developer commentary etc.

The presence of music files doesn't prevent the music from being procedural, the Music Director might sample from that music and stitch it into something new.

Here are relevent quotes from the developer commentary (I think all the music commentary is on the hospital section of the commentary level, I haven't checked the other parts).

Tim Larkin: We took several steps to keep the music interesting enough that the players would be inclined to keep it on as they play. We keep it changing so it won't become tedious; to this end, we created a music director that runs alongside the AI director, tracking the player's experience rather than their emotional state. We keep the music appropriate to each player's situation and highly personalized. The music engine in Left 4 Dead has a complete client-side, multi-track system per player that is completely unique to that player and can even be monitored by the spectators. Since some of the fun of Left 4 Dead is watching your friends when you're dead, we thought it was important to hear their personal soundtrack as well. This feature is unique to Left 4 Dead.

Mike Morasky: A lot of over-reaching dynamic music systems go to great lengths to organize and control a very expressive art form, often to the point of making the results perfectly 'controlled' but also perfectly predictable. Our Music Director aims for 'planned serendipity'. By designing the music and rule sets to increase the probability of beautifull happenstance and to minimize the probability of inappropriate mistakes, we end up with the highest percentage of musical events working as planned, an nice mid percentage of acceptably artfull mistakes, and very few actual poor moments. If you overdesign the music and rule sets, there are no surprises but without surprises, listeners are quickely bored. Ironically by keeping things simple, the music seems planned; greater complexity just leads to greater randomness and many more poor moments.

Mike Morasky: We based the music on what the player is actually experiencing and not on what we want them to experience. Working from artificial life work our designers had done on Lord of the Rings and the Matrix sequels, we implemented a simple system to examine what's going on in the player's immediate environment, the added the appropriate reactive, scalar rule sets to cantrol the music and its volume levels. Most of the more prominent musical events are thus reactive results from rule sets processing this input — making the musical experience specific to each player. This system also controls the dynamic mix system — another feature unique to Left 4 Dead.

At the very least there is a set of tracks that are mixed to give different feels as well as musical events that are triggered on "witch alerted" or similar events. Though the notes and instruments are not procedurally generated, it would be fair to say that the music is dynamically mixed and arranged.

If you have the game, the music resource files are at:

steamapps\common\left 4 dead\left4dead\sound\music

To me they sound like fragments of the in-game music. With the exception of the non-interactive parts of the game they are all pieces 5-30 seconds long that mixed with other pieces to give the in-game music. The longest parts are the "you are dead", "safehouse reached" and "ending credits" music, those are always identical in-game without mixing.

Another reason the line is blurred is that any content affects gameplay. For example, think of how much effort Valve puts into getting the silhouettes of their characters unique, and how their effort that changes only the appearance of the characters makes the game different by allowing players to identify objects faster.

The distinction is still useful and we still focus on PCG, but the only thing this excludes is something like: "making content manually (that is, the hard way) and then finding a way to generate the exact same thing procedurally." In which case it is more of file compression than any thing else.

Which reminds me of the definition of algorithmic information content…

Perhaps I should put in a page on AIC. Will take a bit of research first.

Just my interpretation.

You may want to have a look at the gallery module in the wiki syntax for a useful way to display this. However:

Allowed "per-image attributes are: * link - URL or wiki page name (does not work with Flickr images to be o.k. with Flickr terms)

Anyway, first image:

Algorithm in forums.

Great idea. I can't see a top level page for the gallery: I'm guessing gallery is the best link; adding this to the menu bar at the top of the page once we have some images populated would provide another way for people to navigate the site.

So, for each image:

• Credit for the image (if it is not yours)
• Link to algorithm/source code for the image.

I will start.

Edit:
And we should limit the width of images displayed to 600 pixels or so.

Recently, I've been stuck on the idea of procedurally generated storylines. It just seems to me that all these Game Companies could save an enormous amount of money on personnel if the job of creating mediocre game stories was handed over to computers. Or at the very least force them to come up with something better than the algorithms can come up with. :D

As a non-techie, I'll be around more for intellectual curiosity and knowledge absorption than as an active editor, but I may pop up from time to time, depending on the circumstances.

Cheers!

Starting with a black pixel, the following function is applied to each pixel:

 #value is the value of the current pixel, 0 to 255 #n_per_level is how many subdivisions per step #x is a random value between 0 and 1, straight out of the random number generator def foo(value,n_per_level,x): v = value/255. if v > 1./n_per_level**2: v = 1+math.log(v,4) return int(255 * v + .5) else: return int(255 * x + .5) 

These two steps are repeated until the image is large enough. I cropped the output image to yield my desktop resolution.

Possibly ontogenetic and teleological tags could be added as well.

This is what I have so far:

Map generation

A map requires a set of values over a 2d or 3d grid, it forms the space that the player occupies and interacts with.

Artificial Life
Cellular Automata
City Generation
Dungeon Generation
Dynamic Weather
Fire Propagation
Fluid Dynamics
Forests
Fractal
Fractal River Basins
Fractional Brownian motion
Genetic Algorithm
Height Maps
L-System
Mazes
MegaTexture
Perlin Noise
Plant Generation
Procedural Sky
Procedural Spooling
Procedural Texture
Rapidly-exploring random tree
Pseudorandom Number Generator
Random Number Generation
Reaction-Diffusion System
Texture Synthesis
Universe Generation
Voronoi Diagram

+Sequence generation

Sometimes you need a group of things with a linear order to them, whether notes in a song, letters in text, or numbers as parameters to other algorithms.

Algorithmic Balance
Artificial Life
Automatic Game Design
Blum Blum Shub
Cellular Automata
Dynamic Weather
Fractal
Fractional Brownian motion
Genetic Algorithm
Linear Congruential Generator
Mad Libs
Markov Chain
Mersenne Twister
Natural Language Processing
Procedural Animation
Pseudorandom Number Generator
Random Number Generation
Random Seed
Reaction-Diffusion System
Shuffling
Wave Propagation

+Uncategorized:

I don't know how to group these things.
Hash Function
Identify System
Magic Crayon
Permadeath
Teleological vs. Ontogenetic

Shouldn't this page also include links to various algorithms and stuff like that? There's no central place to find stuff about map generation, links are in here, but other pages are dispersed in algorithms :P — Emile

The articles page was initially a dumping ground for all the external links - but feel free to add in any articles on the wiki itself… don't forget the pcg-algorithm category is another good place to put these.

Suggestions are welcome as to how to get the PCG wiki added as a reference on the Wikipedia page. A naive attempt (adding it) gets caught by Wikipedia spam bot detection. A more complex attempt would involve getting someone to override this.

Any other ideas?

Andrew

I'm a 34 yo mediocre programmer from northwestern Italy…I'm not very sure my contributions will be useful for the community and I don't see myself stopping by in the forums very often, but I saw there are few members registered and thought it was a nice boost anyway to have some new guy around.

By the way, I'll be devouring all the informative material presented here and hope to learn something even if in the end I won't be able to create a "product" (I'm much for studying for pure knowledge).

My actual "dreams" would be applying PCG to world creation, focusing on caves (maybe based on hexagonal-cell mazes) and cyber-megalopolies

Bye all!

So close! If only tags were defined in the page markup instead of that little widget… Hm.

Pros: Structures the content for clear navigation, makes it easy to collect similar pages together for 'index' pages.
Cons: Makes creating links unintuitive, enforces particular conventions, doesn't account for pages that are more general or multifaceted.

I think, however, I have found an easy solution: Tags.

ListPages

Wikidot has an awesome module: ListPages. This does one simple thing with tags that I have longed for on DOZENS of other wiki engines: Allow list of tagged pages with and/or/not.

Basic "OR" usage:

 [[module ListPages tags="game level-generation"]] 

Basic "AND" usage:

 [[module ListPages tags="+game +level-generation"]] 

Basic "NOT" usage:

 [[module ListPages tags="game -level-generation"]] 

Consider a game that uses a cellular automata for fire propogation. It could be taged with: "game rpg cellular-automata fire-propogation". And also an article that discusses this method: "article cellular-automata fire-propogation".

From this we could create several interesting automatic indexes:

Games that use CA: tags="+game +cellular-automata"

Articles about fire: tags="+article +fire-propogation"

All CA pages that aren't about fire: tags="+cellular-automata -fire-propogation" (Not sure if this is hugely useful, but hey, it's there!)

And then, of course, obvious things like All Articles: tags="article".

Why?

Tags are very natural to create, trivial to edit and mash around, fairly easy to maintain. This way also all pages can be in a flat structure, making linking more straighforward, obvious, and constant. As well, the complexity of the system can grow (or shrink) as needs change.

Thoughts?

Graham here — game designer, been in the insdustry a couple years. Teaching myself how to program, mostly through procedural generation and generative art… This site is like magic for me! Thanks huge!

I agree strongly about the namespaces thing; having started several wikis of my own, I have a strong distrust of namespaces. Consider their purpose: to prevent naming collisions. But how many games are going to be called 'runtime-random-level-generation' ? Not many I assure you. ;)

If there are collisions, wikipedia-style disambiguation pages still encourage accidental linking, and increase user exposure to content, rather than hiding it away.

I haven't ever played with Wikidot, don't know what kind of control it offers for rearranging these things (going to start a sandbox and play around with it). This seems like a community I can really get into, I'll report back here if I find anything interesting. :)

• 1,903

Visits

* 1,071
Absolute Unique Visitors

* 9,168
Pageviews

* 4.82
Average Pageviews

* 00:04:01
Time on Site

* 25.70%
Direct Traffic

* 34.63%
Referring Sites

* 39.67%
Search Engines

Sources

Visits

% visits
google (organic)
719 37.78%
(direct) ((none))
489 25.70%
roguelikedeveloper.blogspot.com (referral)
314 16.50%
forums.tigsource.com (referral)
115 6.04%
216.239.59.104 (referral)
39 2.05%

procedural generation
57 7.55%
procedural dungeon generation
51 6.75%
procedural content generation
39 5.17%
procedural content wiki
31 4.11%
spelunky wiki
23 3.05%

* 53.86%
Bounce Rate

* 53.70%
New Visits

category-pcg-algorithms
category-pcg-games
pcg-algorithm:dungeon-generation
category-pcg-software
articles
pcg-algorithm:city-generation
system:recent-changes
list-of-pcg-games-by-genre
pcg-games:subversion
what-pcg-is
pcg-taxonomy:runtime-random-level-generation
the-death-of-the-level-designer
list-of-pcg-games-by-platform
pcg-algorithm:teleological-vs-ontogenetic
major-pcg-games
pcg-games:spelunky
pcg-software:cityengine
pcg-algorithm:fire-propagation
forum/t-60537/introduce-yourself-here
forum:start
list-of-pcg-games
pcg-games:borderlands
forum/c-42075/general-discussion
pcg-algorithm:cellular-automata
pcg-taxonomy:procedural-puzzles-and-plot-generation
pcg-algorithm:algorithmic-balance
pcg-algorithm:blum-blum-shub
pcg-algorithm:fractal
event:tig-source-procedural-content-generation-competition
start
pcg-software:werkkzeug
pcg-algorithm:l-system
pcg-games:angband
pcg-games:far-cry-2
genre:roguelikes
pcg-algorithm:perlin-noise
pcg-games:unangband
pcg-algorithm:artificial-life
pcg-algorithm:forests
pcg-games:infinite-mario-bros
pcg-algorithm:procedural-sky
forum:recent-posts
pcg-algorithm:markov-chain
pcg-links
pcg-algorithm:dynamic-weather
pcg-algorithm:linear-congruential-generator
pcg-algorithm:fractal-river-basins
pcg-algorithm:universe-generation
pcg-games:dwarf-fortress
pcg-algorithm:permadeath
pcg-software:dire-press-random-dungeon-generator
pcg-taxonomy:dynamic-world-generation
pcg-algorithm:fluid-dynamics
rescue-on-fractalus
pcg-algorithm:rapidly-exploring-random-tree
pcg-algorithm:procedural-texture
pcg-software:biome
pcg-algorithm:heightmap
pcg-algorithm:magic-crayon
pcg-algorithm:megatexture
pcg-algorithm:voronoi-diagram
pcg-algorithm:maze
pcg-games:elite
pcg-games:spore
pcg-software:smoke
news
pcg-algorithm:plant-generation
pcg-algorithm:wave-propagation
pcg-games:kkrieger
pcg-software:world-construction-set-visual-nature-studio
pcg-games:rogue
pcg-software:euphoria
pcg-algorithm:identify-system
list-of-pcg-games-by-license
pcg-games:beneath-apple-manor
pcg-taxonomy:procedural-generation
genre:role-playing-game
pcg-games:mount-blade
pcg-software:the-procedurality-engine
pcg-taxonomy:instancing-of-in-game-entities
randomness
examples
pcg-algorithm:pseudorandom-number-generator
pcg-games:toejam-earl
pcg-algorithm:texture-synthesis
pcg-games:artificial-nature
pcg-games:civilization-iv
pcg-software:crowd
pcg-software:fxgen
pcg-algorithm:genetic-algorithm
pcg-algorithm:hash-function
pcg-algorithm:mad-libs
pcg-algorithm:natural-language-processing
pcg-games:apple-tree-feed-the-hungry-birds
pcg-games:infinity:the-quest-for-earth
pcg-games:x-com:ufo-defense
pcg-software:city-maze
pcg-software:world-machine
admin:manage

With regards to the name spaces - it seemed a fast way to build up links between various pages, particularly the classification of games, software and algorithms. You are right in that it's a barrier to entry though.

I guess setting up categories is the 'correct' approach - is that too much work now?