What's Killing Game Development?

General Discussions
#21

Well, in many cases yes. But, believe it or not, I think the “fun factor” can be added into the algorithms. Anything can be created procedurally. You just need to discover what is it about the game that makes it fun, and then somehow import it into the algorithms.

I don’t think games will get smaller, simpler, cheaper. There will always be this demand for large games like Half-Life, and the effort that goes into creating it will always increase… despite of procedural algorithms or better editing technologies.

#22

It’s not that easy to write an algorithm to be “fun”, you can do a lot with genetic algorithms but the one of the most difficult tasks is the fitness function that determines how good a possible solution is.

#23

I’m a fan of understanding what works well… So, what is the example (pattern?) that Nintendo uses to make a profit?

#24

Easy:

  • Cheap machine
  • Cheap games
  • Fun and different controls
#25
  • Highly polished
  • Massive amounts of playtesting and tweeking.
#26

Fun and different controls : not necessarily.

Cas :slight_smile:

#27

What about content? “If you build it, they will come” (?).

#28

Aha. Ahahaha. Hahahahahahahahhaha!

Well, you know - when you find out “what it is about the game that makes it fun”, and you decide to protect that knowledge, and sell shares in the company, count me in for a few thousand shares. We’ll all be billionaires soon enough…

Seriously - you’ve just taken the most difficult thing in the entire games industry, the Big, Unsolvable, Problem, and calmly said it can be done … just like that :).

#29

That’s not normal.

#30

You misintepreted me. I said I believe anything can be created procedurally, that is in algorithms. If you have a fun game, yes, you can recreate it in algorithms. What are games? Isn’t a game code just one big complex algorithm?

#31

Well if something can be created procedurally then yes it can be done with an algorithm; but a game is not just “one big algorithm” it is actually more of a simulation of an imaginary world created by the game designer. The models are created by artists, and sure you can use algorithms to create random textures and random arrangements of buildings to generate scenes, but anything requiring “art” and the concept of good and bad requires much more complex AI and methods.

If you could create an algorithm that decides whether something is “good” the same algorithm will play a pivotal role in an algorithm that makes something “good”!

#32

Isn’t that like you only need enough time, monkeys and typewriters until one writes a nice poem?

Chances you can fit ‘fun’ inside an algorithm in your lifetime, are next to zero.

Back on topic: look at your average building in your average town… it aint exactly fun inside, so why would we aim for ‘fun’ inside our proceduraly generated cities anyway, it’s about recreating reality, and reality aint exactly fun (eh…) well, you need to have a goal to have fun, and the goal has probably little to do with the buildings around you, it’s about interacting with others inside that city.

#33

[quote]Chances you can fit ‘fun’ inside an algorithm in your lifetime, are next to zero.
[/quote]
not entirely true. if you know that a shooter level is fun if it has many alleys and corners, but with some open spaces, you can make the algorithm to do just that :slight_smile:

#34

Well try this for size; read back issues of the Game Developer magazine and you’ll find an article comparing the timings of character responses, level sizes (in terms of how long it takes to get from one end to the other) of all the most popular and highest rated games and they are all very similar. Yet amongst these results countless games with similar timings are not fun to play, in fact how many bond games have come close to comparing to the gameplay of Goldeneye cough timesplitters!! Why is MGS so great to play while Mission Impossible is not so highly rated?

People get it wrong so many times because “good” is not that easy to achieve. And until “good” can be measured or defined to quantifiable means it cannot be procedurally created!

#35

I think I need to reword my sentences.

What I DON’T mean is that you can have a algorithm like: goodGame = algorithm(game parameters)
I think people think that’s what I mean, but it’s not.

What I mean is, like keldon85 said, many levels in FPS games are very similar (the most popular ones). By comparing all the popular maps you can find out some common parameters, and from it you can create similar levels using an algorithm:

goodLevel = algorithm(level parameters)

I have experienced this in a game called Battlezone 2, but I’ve created about 10 popular maps for that game and all I did was following simple rules I had figured out regarding playability, but I pretty much knew what made the maps popular. I could probably write a algorithm to create heightmaps that are fit for that game.

If I remember right, some of the C&C had some random map generator, but it was entirely random and it did not even take into account space for base building, nor some neutral grounds waiting to be taken over. So those maps were no fun to play, but making the maps fun could be easily done by simply allowing for base areas, neutral grounds, and not scatter resources in weird places.

If you can generate whole levels (fun ones too) using an algorithm, and you can generate models, such as characters, I don’t think there is any limit. You can create music by using an algorithm! You can create textures by using an algorithm! You can create trees using an algorithm! You can create buildings using an algorithm! Of course all algorithms need INPUT.

That doesn’t mean you’ll have an algorithm that does everything for you, what I mean is you’ll have algorithms that help you create that good game you’re trying to make.

#36

Those parameters you’re talking about are most likely very high-level, and extremely hard to put in numbers.

Your observations resulting in those values of parameters from other maps required a brain to analyze and turn into ideas about what exactly are the parameters, what values they have, and what influence one has over the other. Using that information to build something new, requires the very same brain, to solve all kinds of problems that you observe while building the new map.

Think of it like this: terrain (valleys, hills, mountain-ranges) is one of the most elementry parts of a map. We, humans, can immediately see whether it was generated by an algorithm or ‘produced by nature’. Thousands of developers have worked on this, and the result is still quite poor. No matter what parameters you feed this algorithm, at best, it will still require a human-brain that determines whether is ‘looks realistic’. Personaly, I’ve yet you see a convincing ingame scenery, that didn’t use actual elevation data.

What I’m trying to say is: even seemingly easy to reproduce structures, are extremely hard to fit into an algorithm. There are so many variables that have to be finetuned (and influence eachother), that the actual INPUT is so big it would be impossible to handle.

Creating fun in procedural algorithms, requires intelligence (being able to analyze and make corrections on the fly). Either that, or tune a gigazillion variables yourself, which takes the ‘procedural’ out of the procedural algorithm.

#37

Without trying to sound like a broken record, it ain’t that simple. Certain types of battle grounds are easy to generate, I mean random level generators are simple to build but don’t confuse a random battleground generator with a game level generator.

Basically if the range of possibilities are range and the need for variation is high but there is no measure on your solution then you have a tough (or impossible) implementation. Buildings, characters and so on are easy to generate since there is a model that they can follow - I think you’re putting too much thought into the idea that algorithms can create anything.

A good level is not just level parameters in the same way that music or any other form of aesthetics is not just parameters. Read up on AI and genetic algorithms and remember that the fitness function is half of the key. Without a working fitness function you have no genetic algorithm; and then there is the gene types, gene mixing approach and so many other areas to get right - but other than that all genetic algorithms have the same make up.

Some things have few rules, I mean what makes a good battle ground in command and conquer, what makes a good balance of levels under different difficulties in a game like Goldeneye. One more thing about a game is that the levels are not just stand alone components, Goldeneye worked because it had the perfect first level, it had a perfect follow up to it, and right through the different types of levels followed on from each other to give a good variation of playing styles. Time splitters and Perfect Dark had a similar formula to that, but the formula only works with a full understanding of what it feels like to play.

There is a reason why games are play tested, it’s because the work is not finished straight after the design phase.

But again read up on AI and genetic algorithms, it might give you more insight into the complexity of “good”.

#38

Well, its a general rule in game design that your first level has to be completely overwhelming yet not difficult, so its not a complete surprise. Looking at games like God of War which excess this in full, never played the rest of the game but playing the first level just so asks for more. Ofcourse you cant apply procedural level generators to that kind of thing, at least not to full extend, but im sure you can find some application for procedural content creation so that it takes time of your hands.

I guess there are three main things you have to concider when using procedural content; first, it has to lower development time and cost (thats why we are discussing it in the first place ;)), second you have to look where you can apply it, how and in what way, and third you have to be aware that its not simple and easy to just come up with some kind of algorithm that just makes it all happen.

#39

Procedural graphics would help you create the levels/maps you want in your game, it wouldn’t create the levels/maps for you.

It’s a question of; a) starting from scratch, creating every single model and texture or b) having algorithms creating the generic things for you so you can spend more time designing your game.

You can pretty much create an algorithm for anything:

  • tables
  • chairs
  • light pole
  • rocks
  • trees, all types
  • etc.

Once you’ve created an algorithm for, let’s say table, you can simply click a button in some level editor, define some parameters like what the material the table is made from, how many legs, the dimensions, if it’s round or rectangle in shape etc. etc. and bamm, the table appears without any work from the graphical department!

For your next game title, you might want some fancier table, so you simply add bump mapping feature into that same algorithm to give it some rough look.

IMO we’ll see features like that.

All real world objects can be represented in math/geometry.

#40

Someone’s still got to make all the textures.

Procedural isn’t the future, it’s just one way of achieving a particular effect (seemingly large amounts of bland content).

Cas :slight_smile: