next »

[Aquanoctis]

Yep, that's the idea.

[DavidCaruso]

here's something to brave: procedurally generate a good game

Experiment: Spend a shitload of time making 500 or so expertly crafted levels. Make it so that the game randomly picks a level each time a person loads up the game, and that whenever a level is first loaded a flag is set server-side so that no person will play the same level as another person on any computer if at all possible. Do not release this game in public (because it would quickly reach a point where new levels couldn't be chosen and old ones would have to repeated, and someone might figure it out), but instead submit it to IGF judges and claim that you created a master AI that uses neural networks to simulate Warren Spector's master game designer thought processes and procedurally generate levels in realtime (make fake loading screens). Win technical achievement award.

[gimymblert]

By the way, the fun things is that AAA use procedural generation everywhere!

[e-dog]

I wrote an article about water erosion on heightmap terrain. Code included.
It's more about modifying existing data than generating new one, but I think it still can be useful to people here. I figured a lot on my own since internet was lacking detailed info on the topic.

[Theophilus]

I wrote an article about water erosion on heightmap terrain. Code included.
It's more about modifying existing data than generating new one, but I think it still can be useful to people here. I figured a lot on my own since internet was lacking detailed info on the topic.

That is just plain beautiful.

[JobLeonard]

Dammit, the website seems to be down..

EDIT: Seems to work again, reading. Looks really interesting

[e-dog]

Dammit, the website seems to be down..

Huh? Try again. For some reason it failed for me too, but worked on the second try.

[wademcgillis]

A little blog post on how my poorly made procedurally generated "caves" are made, including full code (AS3).

http://wademcgillis.com/?p=560

[Momeka_]

Wrote a blog post about a way to generate some simple maps with a maze algorithm. Includes source code.

[sublinimal]

Found an interesting blog post on Reddit: http://accidentalnoise.sourceforge.net/minecraftworlds.html

[DeadPixel]

Here's an interesting article on herringboned Wang Tiles that was linked to me tonight.  I think I'll try an implementation of it for my current game and see what I get.  http://nothings.org/gamedev/herringbone/

[Ashkin]

Just finished my LD entry, which used kind of procedural generation. AND NOW I SHARE THE PROCESS.
Step 1: Create a blank array of cells, each 9*9 tiles. This example uses 5*5 cells.

Step 2: Scatter points around randomly for the player, block, and kiln to spawn. Add an extra 'dummy' point for taste. Mark each of the cells these points are in 'visited'.

Step 3: Draw lines from the player point to the block point, and the block point to the dummy point and kiln point. Mark each of the cells these paths touch as 'visited'.

Step 4: Carve out the 'visited' cells into empty rooms. Note down exits to each room.

Step 5: Check the available exits in each room. Choose one configuration at random from five possible configurations for that exit configuration (75 room types total). Place these rooms into the correct cells.

Step 6: Output all these to a tilemap and you're done!
(Excuse the tiny image size. One pixel = one tile.)

[LuisAnton]

Most entries in this thread talk about procedural terrains or dungeons.

What about puzzles? Given a set of pieces and rules, like chess or physics, or mario bros (http://www.supermariobrothers.org/infinite-mario.html), which techniques can be applied to create riddles to be solved by the player? It depends a lot on the game, for sure. Have you achieved something similar? Random levels for Angry Birds, perhaps?

[Geeze]

I think puzzle generation is less covered because simply generating good and fun puzzles is usually nearly impossible.

[st33d]

Start at the solution and then execute a number of moves away from the solution.

The only problem is that it would be hard to judge the difficulty.

[vinheim3]

Not only that, but the puzzles would be very game-specific, so it would be hard to relate to. However, it would be interesting to see how people randomly generate their puzzles based on their game concepts.

[rek]

If you know all possible moves and the overall structure the level has to take, you can come up with a PG system to produce it. In theory, you can also make a system to test it (even if just by trying random things) until the goal is reached, then keep/reject the level accordingly.

[e-dog]

Start at the solution and then execute a number of moves away from the solution.

The only problem is that it would be hard to judge the difficulty.

Actually, this works quite well for Mahjong and the likes.

[JasonPickering]

you could take a look at Block Rogue.

[LuisAnton]

you could take a look at Block Rogue.

I would love getting an insight into those "ingenious and sinister algorithms" :D Sure enough, it depends a lot on the set of tools at hand. I might explore that line with some toy puzzles...