When is Random Too Random?

[Pueo]

    At what point does a game become too random?  Of course, random map generation and item/monster placement is pretty much necessary for a roguelike, but when do you stop?  Should the items themselves be random?  Should monsters have random stats and powers and movement patterns? 

    What about combat? Is it a bad thing if you know goblins always have 75 health? 
 
    Another part is setting and narration.  Should quests be random?  Would a random setting make the game too confusing?  If the game has a story, how much of it should be constant?

    I feel that past map generation, item/monster placement, and item generation, it becomes too incoherent.  It becomes too hard to gauge the strength of monsters; even if you know it's a goblin, it could have anywhere from 50 - 100 health (for example), meaning you might charge in, thinking it'll go down in a couple hits, but in actuality it takes 5, nearly killing you in the process.

[kraflab]

This is something you really have to decide for yourself, because there isn't an answer really.

Imagine a spectrum with Puzzle on the left and Roguelike on the right and decide where you want your game to be.  Ideally, a roguelike is like a randomized puzzle where the strategy comes from reacting to unexpected events and dealing with what you find.  If it's too random then there is no strategy, and if it's not random enough it's too predictable and becomes less fun.

Another thing to consider is how much information the player has.  Again I think this comes down to a puzzle-roguelike spectrum.  On the puzzle side, the player knows everything.  They are omniscient and the game is about figuring out the best/fastest way to complete something.  On the other side, it is the lack of knowledge that creates a meta-game for the player, where learning which monsters do what and experiencing surprising effects are part of the fun.  Taking risks is an important part of many roguelike games, and it lies in the fact that the player-character cannot be sure about certain things (i.e. they can miss, get crit hit, etc / do I use a potion this turn or next turn?).

You can exchange these things without moving on the spectrum.  I.E. you can give the player more information and also increase the randomness, such that perhaps a goblin now has 50-100 health, but the player can view enemy hp.  This keeps things random and allows for surprises, but doesn't leave the player scratching their head when an enemy refuses to die.

It really is up to you, and different people will like different positions on these spectra, so you should do whatever you want to do.  For instance, Microgue has randomized levels where the player is omniscient, while most roguelikes leave the player heavily in the dark.  You can also mix and match, for instance having certain enemies and abilities that are heavily deterministic.  An example would be an axe that does 2d6 and a sword that does exactly 6 damage.  This also helps to make weapons and monsters more interesting.

You can make any gameplay mechanic fun, it's all in the implementation (don't listen to DGrey )

[Paul Jeffries]

It depends what you are trying to achieve.  Generally the aim behind randomisation is to prevent knowledge of previous games from spoiling various discovery-related aspects of the gameplay (i.e. the map layout, item identification etc.).  But that isn't the only thing you can use it for.  For example: I think POWDER does the thing you are talking about where it slightly randomises the strengths of monsters.  I think the aim there (I'm guessing) is to keep the player alert and stop them from treating enemies as 'windshield bugs' - even if you're facing a monster that is much weaker than you generally you still need to be careful and can't just hold down the arrow key until they disappear with no real risk, because you can't be entirely sure how strong they actually are.

I think randomisation only becomes 'too much' when it throws up situations where whatever the player does they have no chance of winning.  I've been playing Hack, Slash, Loot recently and this suffers a lot from this problem: how well you do in a given game is almost entirely down to what items you come across.  This is not so much a case of there being too much random stuff, however - it's more that there's not enough other stuff in there to give the player the ability to deal with that randomness: you have to use/equip items as you find them and there's no other kind of skill system - so the only real tactical element you have control over is positioning and that doesn't have that big of an impact on combat compared to whatever items you happen to have randomly come across.

[Pueo]

@Paul, I played (the demo of) Hack, Slash, Loot, and I think you're right.  Two things are random that make it insane to play:

A: From what I can tell, your character is blind in one-and-a-half eyes and has zero muscle coordination.  Apparently, it takes 10 tries to hit once, and then it's usually a one/two-hit kill.

B: Good artifacts are very hard to come by, making it less a game of exploration than a game of hunt-for-a-weapon.

@kraflab, That slider between puzzle and roguelike is so finicky!  I don't want players to be omniscient, but I do want them to be thinking about the "smartest" way to do something.  I want to reward tactics and make them preferable over engaging in open combat.  I don't, however, want the game to be the same every time.  A few things will be the same (the types of monsters, for example), but I'm trying to figure out how much needs to be different to make it interesting every time.

[guest509]

  As far as randomizing monsters I think it's not biggie if the same monsters have the same stats every time. This allows players to get better at your game, as they know what they are up against.

  The variety and replayability comes from not knowing when or what monsters you'll be faced with. Also throwing in the odd 'out of depth' monster can be fun. Fate likes to throw in the odd 'elite' style monster that is like the normal version but tougher, or has a random ability or immunity.

  So I guess I'm saying most games that do it really well are those that toss a curve ball once in a while.

  Random Realms randomizes monster stats and names each game, and introduces a new monster each level. So level one will give you an 'A' to fight, level 2 a 'B' and so one. Between levels you get a heads up about the monster you'll fight next. I'm not sure if out of depth or elite special monsters are used in Random Realms, but it's the game I think of when guys want to talk about randomized enemy stats.

  As far as randomizing item stats, that can be fun. Like giving each mundane item a chance to have an enchant and occasionally an 'artifact' will have several enchants. Random enchants each time. The issue here, obviously, is you might be randomly creating complete junk.

  There was a 7DRL that did something along these lines this year. Well of Enchantments? I didn't play it, but each item (and creature?) was created from the same class. Maybe check it out. If you are pretty code savvy the source code would be neat to read. I'm not so code savvy, other people's code makes my eyes bleed. :-)

  As for randomized quests, or even a randomized win condition, I point to the game Fate. You should really check out that game. The original game is best I'd say, you can find it cheap or even free somewhere. It's a single dungeon, single starting town, randomized item stats, randomized main boss enemy, randomized quests. No classes, you can gear up and stat up anyway you like. If you beat the game you can hand down an heirloom item to your next character, but you don't even need to quit when you beat the main boss. You can keep going, it'll keep making harder and harder monsters and what not. It's not a perfect game, but it has a lot of interesting qualities.

  Randomized quests seem totally doable and interesting to me, it's just tough to make them interesting. Side quests are rarely interesting anyway, I find making the END GOAL of the game different each time to be very very interesting. Slay the Dragon this time, you need to load up on fire resistance. Slay Dracula next time, better find a lake of holy water to dump him in. The next time it's a Hydra, blunt weapons are good, but a scroll of multi-fireball is the real winner.

[Pueo]

 So I guess I'm saying most games that do it really well are those that toss a curve ball once in a while.

 Yes, I think for something to be interesting it has to be unpredictable (to a point).

 If you are pretty code savvy the source code would be neat to read. I'm not so code savvy, other people's code makes my eyes bleed. :-)

  I'm probably the least code savvy person on the forum.  I mostly dream about how awesome my roguelike would be if I could just get to learning (and I mean really learning, I have basic knowledge of C) a language.

 Randomized quests seem totally doable and interesting to me, it's just tough to make them interesting. Side quests are rarely interesting anyway, I find making the END GOAL of the game different each time to be very very interesting. Slay the Dragon this time, you need to load up on fire resistance. Slay Dracula next time, better find a lake of holy water to dump him in. The next time it's a Hydra, blunt weapons are good, but a scroll of multi-fireball is the real winner.

 My original goal was: Have 20-ish monster types, have a boss type of each of those, so 20-ish bosses.  Dungeon dive to about depth 25-ish, fight random boss (out of the pool of 20-ish bosses).  Go deeper to fight the rest of the bosses, or high-tail it out of there for your reward!  Unfortunately the whole needing-to-know-how-to-computer-program-thing got in the way of designing, so now it's just a fun idea.
   Now that I think about it, I have tons of "fun ideas" that I can't really implement (again with the computer programming ).  I guess I just have to choose one language and stick with it; I keep bouncing around because I think one is "better" or "easier" than another, but they're all good in their own ways.  I'm looking at Haxe now.

[guest509]

  I have the same problem. I bounce around to different ideas. I made a thread once about how many design choices are made because doing it the way you really want is just too hard to program.

  Lol, I got responses along the lines of, "Well even first person 3D shooters are one man projects now." Very useful info! Not.

  So I think it's about picking reasonable goal and sticking to it.

  As far as learning a language, you can't go wrong with C++. Pretty standard, and after you learn that the rest is easy. If you are savvy you can be kicking ass in that language in a year.

  I gave up on being a hard core programmer, though I worked in C++ and Turbo Pascal for years in school. I use Gamemaker now. It has recently seen a TON of advancement, can put out HTML5 (browser) programs and apps that will run on mobile devices, mac, windows, whatever you want. The price has really gone up though. So that sucks.

-Jo

[Pueo]

  I have the same problem. I bounce around to different ideas. I made a thread once about how many design choices are made because doing it the way you really want is just too hard to program.

  Lol, I got responses along the lines of, "Well even first person 3D shooters are one man projects now." Very useful info! Not.

  So I think it's about picking reasonable goal and sticking to it.

    I feel the same way.  I jump around a lot with video games as well (I enjoy thinking up ideas based on games). 

  As far as learning a language, you can't go wrong with C++. Pretty standard, and after you learn that the rest is easy. If you are savvy you can be kicking ass in that language in a year.

    That's why I'm looking into HaXe; it can convert into C++ source code and several other formats.

[kraflab]

Epilogue has monster variants such as kings or champions with either higher stats, better weapons, or both.  This enable the inclusion of harder creature varieties without leaving the player in the dark (it is very clear when you are fighting a variant).  I also use a split between common monsters and rare monsters, with the rare ones being a bit tougher.  Early on in the game, the magnitude of the differences isn't too staggering, while late game enemies can really pack a punch.  This gives players a chance to beat up the tough guys until things start to get real.  Traveling through the same old floors and finding an enemy you have never seen before is a great experience for me, so that kind of randomness really helps with replayability I think.  Heck, there is even an optional boss floating around that most people will never find

[requerent]

I've been thinking about this topic a lot lately.

I'm a fan of all game-content being generated, but it should be done at pace that the player can react to.


I was considering a 3d pcg game in which 8 or so drastically different isomorphic character meshes could be interpolated between one another to produce an infinite array of monster meshes (UVs in SVG could undergo parallel interpolation to result in a bitmap that corresponds with the new mesh, then apply different regions with pcg textures and... bam- mesh+skin).

Now- the more interesting component to this idea revolves around how gameplay is generated. Choices are divvied up into powers. Powers give a player some number of abilities. These abilities are combinations of a set variety of potentially interesting interactions. For example- there might be a 'pull' power and it might be generated under the limitation that it can only be used at some numeric range. Given any number of possible effects, generating a new one based upon a variety of restrictions, combinations, and sequences can create a lot of cool powers (which would probably be culled using genetic algorithms in quickly simulated combat to ensure that the power is pertinent to the area- literally, naturally select new powers). Combat revolves more around stressing your opponent's powers out (by pursuing relative advantages) until they are disabled- once a creature has no powers it's effectively dead, and you can take its powers and replace them with existing ones (or combine powers using binary shifting or something thereof).


The point-- the powers themselves would have associations with different meshes and textures. Creating visual correlations allows us to interpret what sort of tactical decisions we need to make.

The only other part of it is determining when to add new variants. If you add them rarely or ensure that only one is added at a time, it gives players opportunities to adapt and think about things- especially if those variants generated under winnable conditions. That is, don't add something new unless we know the player can defeat it. This isn't entirely unreasonable if we consider overlapping regions. The weaker creatures of a region will be pushed to the edges. The closer we get to the center of a region, the more powerful the opposition- but it's in theme, so we aren't terribly caught off guard.

Making the game world random, up to the inhabitants and the abilities in the world, is awesome. But it must be done in a way that is coherent.