The World Hates You (mcc)

This game uses an evolutionary algorithm to generate platformer levels. The goal is to create the hardest platformer levels possible (which are still winnable). A central online server is used to breed together the most difficult levels and propagate them to other, future players.

There is no win condition. The purpose of this game is not to be won. The purpose of this game is to get progressively better at killing you.[Author’s description]

[Windows Download]
[Mac Download]
[Linux Download]


  1. Sounds cool, but I’m just getting a blank white screen on my Mac and I have to force-quit.

  2. Interesting idea. Art and audio are fitting. The (seemingly?) inconsistent walljump threw me off, but wasn’t that big of an issue.

  3. This game is, for the moment, fantastic. Very few impossible levels (Tank 7 is hosed, 9 is getting a lot of impossible iterations) and it’s very rewarding when you make it through a really tricky one!

    The specimen numbers are in the 200s now. By the 500s I expect the game will be unplayable.

    • Tank 7 is back in business!

      • That’s kinda crazy because I think tank 7 was busted from the moment I last reset the server! Which would mean it was broken for several tens of generations, then randomly mutated with no evolutionary guidance until it became playable.

        (The fitness function in this game is structured in such a way that no actual evolution occurs unless there is at least one playable level per generation.)

        • Hah, weird. I think 7 is probably the most interesting tank right now. The playable iterations of 1, 3, 4, 6, and 8 all have easy routes up the side of the level, and that’s the big weakness of the evolutionary system: in order to change this, you need a harder route to randomly open up at the same time as the easy route randomly closes down. If the harder one opens first, then nobody uses it and the level is still beaten with a low death count. If the easy route closes first, then the level is impossible so the mutation is thrown out. Whereas an already interesting route with tricky jumps (as in tanks 5, 7, and 10) can change subtly and slowly grow in difficulty, the others are kind of stuck for now.

          One thing that might speed things up is adding a quick is-the-goal-connected-to-the-start-via-empty-space check. That should allow the computer to throw out a lot of the impossible levels immediately and speed up the evolutionary process. In fact, maybe replacing all tiles disconnected from the start with walls would simplify the levels to allow more focused evolution on the parts that a player actually uses.

  4. OMG deceased crab played one of my games! That is seriously exciting to me.

    Buttons– if you don’t mind me asking, what Mac OS X version are you running, and if you open System Profiler and go to “graphics/displays” what kind of video card do you have?

  5. I like the idea for this game, but I feel there are a couple of things holding it back.

    For one, the controls are really bouncy and slippery, which I guess is a matter of personal taste, but the hazards are also bizarrely unforgiving – why does this ( ) kill you? I feel like together these make it easy for levels to be “frustrating” hard rather than “challenging” hard (which seems more like what you’d want), as well as leading to statistically misleading deaths.

    I also feel that the selection process is a bit flawed, in that the first couple of levels you play in a tank will usually be the ones you die the most on, just because you aren’t as familiar with the level. It might even out with enough players, dunno, but more tanks or more variations between levels wouldn’t hurt either.

    But overall it’s a neat idea, and might be worth working into something more polished!

    • I think asking an evolutionary level generator purposely designed to make levels as hard as possible to be “fair” or even “enjoyable” in standard platform game terms clashes a bit with the sensibilities working here.

      This game, I believe, is more about the computer than about the player, and I don’t feel player-centric features such as fairness or challenge really fit here. I mean, the game is built so that it naturally leans toward futility. For the player, that is. But the computer still wins! That’s how this game is a beautiful experiment.

      As to the selection process, I have no idea about how it works internally (I’m curious though) so I won’t make any judgments.