Brilliant Observations by My Brother

[guest509]

My brother is not a highly educated man. He is not a game developer. He doesn't wrap his head around Roguelikes AT ALL and isn't a member of the indie gaming scene.

That said, he is an avid gamer. Plays all the big RPG's that come out. He's perfect for running ideas past because he doesn't see them through the jaded goggles of a developer.

When I first started getting into game design and roguelikes in general the venerable KRICE told me that it would change the way I saw games. It would impact my ability to just enjoy a game with out being critical of it's design. Without wanting to 'fix it' or comment on it.

He was right. I've lost the ability to just enjoy the wiz bang of a AAA game.

My brother, though, is full of awesome feedback.

Like I told him, "Dude I would love a good Zelda game but procedurally generated."
Brother:"So you'd keep the worst part of Zelda and get rid of the best part?"

What he's talking about here is the level structure and careful progression of a Zelda, Metroid, Castlevania style game. And he's right. Zelda's combat sucks. Metroid and CV are great, benefitting from the side scrolling, but Zelda was wonky.

So was Gauntlet. Without straffing the ability to shoot was kinda awkward. Maybe if you let the player strafe if they held down the shoot button. But that control scheme doesn't really transfer well to computer. As much as I want people to play with a gamepad on the computer, we know it happens like 0.1% of the time.

Berzerk seemed to do the top down shooter better. I dunno why, but on Atari that control scheme just felt right. Still, that won't transfer to computer.

One arcade control scheme that does transfer I think is the robotron type control. WASD to move and arrows to shoot. Binding of Isaac does this. I didn't dig that game so much, but that's how they did it.

Anyone out there have an avid gamer friend who only really plays the AAA titles? You should run stuff past them. I think you'll be surprised.

[naughty]

Robotron style controls with WASD for movement and mouse for aiming works really well on desktops (Mutant Storm was a good example), not so great on laptops though.

I'm not quite sure what would be bad about a procedurally generated Zelda game. Surely the careful progression and dungeons as a large puzzle is what you'd be trying to generate. i can see why someone would say the combat isn't great though, compared to Demon's/Dark Souls say.

[Vanguard]

Good arguments can be made in favor of Gauntlet-style shooting, the foremost being that it forces you to choose between attacking and retreating.  If you've got WASD + mouse or twin stick controls or whatever, you can backpedal while you shoot and you don't have to make any decisions.  In Gauntlet/Contra/Pocky & Rocky you have to choose whether you're going to attack or retreat because you can't do both.  It adds a little bit of depth.

The problem with procedurally generated Zelda is that the level design, powerup placement, side quests, and puzzles are the best parts of Zelda games.  Those are all things that human being do much better than machines.

[naughty]

...

The problem with procedurally generated Zelda is that the level design, powerup placement, side quests, and puzzles are the best parts of Zelda games.  Those are all things that human being do much better than machines.

The question is really can we get close enough to handmade Zelda levels to make it worth the while. With tech like this and this it might not be a far fetched as you'd think. I am assuming a 2D Zelda game though, doing it in 3D is far harder.

[guest509]

Yeah I was thinking classic Zelda style. Top down. Breaking it up room by room could work, like in the original, but not necessary.

[requerent]

When I first started getting into game design and roguelikes in general the venerable KRICE told me that it would change the way I saw games. It would impact my ability to just enjoy a game with out being critical of it's design. Without wanting to 'fix it' or comment on it.

He was right. I've lost the ability to just enjoy the wiz bang of a AAA game.

The same thing happens for books, films, screenplays, and lots of other junk. But the truth of the matter is-- it provides a whole new level of appreciation. For the design, all of the work and sweat becomes visible. AAA titles become a marvelous engagement in an extremely complex development pipeline. Mo-Cap and normal maps and different lighting effects and shaders and ... it's just sooooooo much more beautiful when you know what you're looking at!

In games, and even more-so in films, it's common to see an 'orange and teal' color scheme. Once you study lighting and composition, you don't even have to pay attention to scenes-- you can read exactly what the film is trying to accomplish by the color schemes used. Watch the Star Trek movie with 'orange and teal,' in mind. It's basically the entire film.

[Darren Grey]

I find it hard to watch TV after podcast editing. I can tell where they cut and paste things so easily. I know now how simple it is to to take time out and switch things around and make the meaning of everything look different.

With games it's not so bad for me, I can still let myself go a lot. And when a really well-designed game appears I enjoy it all the more - I can see how everything works together.  Developer intention is a great thing to see when it works well.

[naughty]

Yeah I was thinking classic Zelda style. Top down. Breaking it up room by room could work, like in the original, but not necessary.

One of the issues with trying to create something approximating a hand made level is that we don't really have a robust language and analytical framework to talk about level design. A lot of hand-made levels are made in an interactive process where people know when the level is good but not necessarily why it's good. Without a theory of level design it's impossible to really expect any procedural algorithm to be meet the standards of hand made levels.

For example there's been mention of dead-ends and loops in roguelike levels design. Mainly that dead-ends without a reason are boring and cause tedious backtracking. Loops are good because they give options for escape and exploration (and reduce the chance of dead-ends) but if you have too many of them it can also be a problem. If we could measure 'loopiness' we could control for it.

There are some interesting articles that do analyse games level design though which could be a start. Here is one for Zelda and here is one for Super Metroid.

[miki151]

If we could measure 'loopiness' we could control for it.

That shouldn't be too hard.

[Vanguard]

AAA titles become a marvelous engagement in an extremely complex development pipeline. Mo-Cap and normal maps and different lighting effects and shaders and ... it's just sooooooo much more beautiful when you know what you're looking at!

Pity about their gameplay though.

[Quendus]

If we could measure 'loopiness' we could control for it.

That shouldn't be too hard.

Indeed it's not. http://en.wikipedia.org/wiki/Connectivity_(graph_theory)

Connectivity is a measure of the number of "alternative paths" between points in a graph (a set of points and connections between points - essentially an abstract description of a roguelike's map, or of its room-corridor diagram).
If it's 1-connected, then the map contains a pair of points that you can disconnect just by cutting off one connection.
If it's 2-connected, then the map contains a pair of points that you can disconnect by cutting off two connections (and if you just cut off one connection, there's always an alternative route between any pair of points).
If it's 3-connected, then the map contains a pair of points that you can disconnect by cutting off three connections (and if you just cut off one connection, there are always two alternative routes between any pair of points).

You can probably calculate the connectivity number with the min-cut algorithm in Boost. There are some caveats, though - just being able to measure the connectivity doesn't mean being able to control it. If you have a minimum and maximum connectivity for a level to be accepted, you need to be able to regenerate the level reasonably and compute the connectivity cheaply, or you need functions that can add or remove paths to increase or decrease the connectivity.

If all your corridors are two tiles wide, the reported connectivity will be different depending on whether you input the room/corridor graph or the graph of the whole map grid. if you have a mix of two-tile and one-tile corridors, the value will be difficult to interpret.

It's also a very strict measure. If you have a level with connectivity 3, and you add a single dead end, the resulting graph has connectivity 1.

[Quendus]

So essentially connectivity lets you measure how many scary monsters you can drop in a level while still giving the player freedom to roam without being instakilled.

[eclectocrat]

Interesting that I'm working on metroidvania-ing my roguelike now.

It seemed a lot harder than it turned out. The important thing is to get data that is an actual graph structure and then use graph algos to ensure it works. My implementation is such that I have frequent insanity checks that trigger wholesale regeneration. There are too many edge cases to handle gracefully.

Regardless, I think that non-linear, looping levels are pretty cool. It will be quite some time before a world is generated which stands up to something like Super-Metroid, but I don't think it is out of the question.

...

The problem with procedurally generated Zelda is that the level design, powerup placement, side quests, and puzzles are the best parts of Zelda games.  Those are all things that human being do much better than machines.

The question is really can we get close enough to handmade Zelda levels to make it worth the while. With tech like this and this it might not be a far fetched as you'd think. I am assuming a 2D Zelda game though, doing it in 3D is far harder.

There is also the third option of using a good algo to generate a skeleton which a human can easily edit, and the fourth option of using pre-made sections in randomly assembled chunks (the approach I take with Mysterious Castle for indoor environments).

I love this topic, I think I'll start a dev thread on this later.

[miki151]

I think the basic algorithm for creating corridors with an A* lets you manipulate the "loopiness" by assigning a weight to tiles that need to be dug out. High weight - it will reuse existing paths -> less loopiness, low weight - vice versa.

[guest509]

Creating 'key items' that open up areas by allowing the killing of a boss and/or allowing traversal of terrain is pretty neat.

It's like you almost have to work backward, like so...

To kill the Phantom Boss you have to hit him with the electric staff,
To find the electric staff you need the raft,
To get the raft you need to kill the Bowman,
To kill the Bowman you need the shield,
To get the shield you need coins,
To get coins you have to kill baddies,
To kill baddies you need a sword,
To get a sword you need to find the Old Man.

Etc...