For some genres, procedurally-defined levels are fine, but when handcrafted levels are done properly there's a pretty substantial difference. I'm reminded of an interview[1] with Shigeru Miyamoto and Satoru Iwata discussing the first level of Super Mario Brothers:
"But if you avoid the first Goomba and then jump and hit a block above
you, a mushroom will spring out and you'll get a shock. But then you'll
see that it's going to the right so you'll think: 'I'm safe! Something
strange appeared but I'm okay!' But of course when it goes against a pipe
up ahead, the mushroom will come back.
At that point, even if you panic and try to jump out of the way, you'll
hit the block above you. Then just at the instant where you accept that
you're done for, Mario will suddenly shake and grow bigger! You might not
really know what's just happened, but at the very least, you'll realize
that you haven't lost the turn."
The author makes a good point that randomly generated levels significantly affect the way that a game is played, but they shouldn't by any means be considered superior to handcrafted levels - players have different preferences for playstyles, and game designers should consider which playstyles they intend to appeal to when deciding on a genre and a level design strategy.
Consider the Megaman series - players are encouraged to master each level, and speed runs are a major draw and source of replay value. Similarly, part of the brilliance in the level design in Sonic 3&Knuckles (that disappeared from later Sonic games) was that it encouraged the player to try different paths to find an optimal route for finding special stages. Exploration and speed runs are a draw for the Zelda, Mario, and Metroid series as well - the quality of level design is a major selling point for the core audience of these series.
Randomly generated stages certainly have their advantages for replay value, but they also remove incentive for exploration and speed runs, which means you're potentially forfeiting the audience that prefers those play styles.
To go a step further into the realm of "video games as art", consider the RPGs and survival horror games like Fatal Frame - cinematography is a significant element, as is the player's emotional connection to certain areas. Randomly generated levels don't create as strong a connection with players, which diminishes the designer's ability to use locations as a significant story element. It would be difficult to create such strong storytelling without a high level of control over the environment.
Some genres, like Arcade Shooters, RTS, and FPS may lend themselves less toward these play styles, but it's certainly not restricted by genre, either - Bullet Hell shooters, like Ikaruga, for example, are designed with players who favor level mastery in mind. It all depends on the audience you want to appeal to and the game elements that you want to focus on.
Sonic 3 & Knuckles was a beautifully crafted game (and one of the few popular platformers of the time where there were multiple routes through levels). Its still one of my favourite games to this day.
"consider the RPGs and survival horror games ..."
Speaking of survival horror games, the Left4Dead procedural editing of levels works quite well (though the general layout is handcrafted). I think a similar mechanism would work quite well for games where completely procedurally generated levels aren't desirable.
One problem with Left 4 Dead is that you rapidly run out of levels (even though each level is very well designed).
I think random levels would work very well, especially for online gaming. If a team didn't know the level, it would really feel like exploring a dangerous zombie infested world.
Oh, I agree. I just meant that its a reasonable intermediate level of procedurally generated levels. I think for a future version they could ramp up how much is randomly generated instead of hand crafted.
I like the idea because you get a nice mixture between replayable random levels and nice and well designed human designed levels.
Theres definitely a lot of room for future improvement though and I'm excited to think about what we may have to play with in the future.
> Randomly generated stages certainly have their advantages for replay value, but they also remove incentive for exploration and speed runs, which means you're potentially forfeiting the audience that prefers those play styles.
You can retain the advantages of static levels with randomly generated ones. In the game I'm working on you can specify the random seed used. A high score list is maintained separately for each seed.
I can't stress how fantastic it is being free from the chains of handcrafted levels. You can really iterate and refine the core mechanics all the way up to end of production. It's liberating!
For some genres interesting levels are a must though. My own current project is looking at how to generate platformer levels that are both interesting and possible to beat. I've got a pretty powerful algorithm up and running, with a pre-trailer up at http://www.pwnee.com
I was already kicking around the idea of using genetic algorithms for play-balancing strategy games (scoring the difficulty based on how many generations it takes to find a winning strategy).
It looks to me like you've taken the concept to a whole new level, though. Good job!
I'd love to play around with some genetic algorithms in a game design setting. It sounds like people have been having some success using it in StarCraft, and I think I could definitely find some use for it in my level generator. I might just have to explore that route for the sequel!
Turns out I'm a lazy programmer when it doesn't involve pretty moving graphics. I suppose the old blog deserves some code love too though. Should be fixed up now, thanks!
The insight is that behavior is more interesting and flexible than content.
The fallacy in the title is the idea of a "level" entirely encapsulating the gameplay opportunities. The point is that the game developer needs to trade off static content vs. dynamic behavior. Some "behavior" may look like content.
Or next we'll argue than randomly generated graphics are equally liberating,
I like the opposition of lifelong hobbies to disposable media.
Next time someone asks if games are art, dismiss art as disposable media. But seeing art as a consumer product is sterile. It's healthier to consider it a lifelong hobby. In the end, conclude that the artist is the one who plays the game, not the author.
I wasted so much time playing the first Steambird, it's great, a masterpiece! I really hope the second one won't become my lifetime hobby.
Deeper, more meaningful systems yield lifelong hobbies, not disposable media
I like how he opposes lifelong hobbies to disposable media. It's completely natural coming from a game designer, but to me it also sounds like he unintentionnaly questions the meaningfulness of art. This inspires me sentences that are not entirely clear, late at night.
This difference is plainly visible if you look at games you're still playing over and over again after you've 'won'. I have a copy of Master of Orion that I've been playing since 1991.
Interesting. I was actually thinking about Tyrian as an example of a game where random level generation would work well (in contrast to those mentioned in my above comment), and wouldn't detract from nostalgia replay value. Another example would be X-COM. Both excellent games.
