All Games Are Puzzles

A while back I read Greg Costikyan's excellent essay "I Have No Words & I Must Design", which starts out by defining exactly what a game is not. For instance, a game is not a puzzle. Puzzles are static, while games change with the player's actions.
I was later contemplating a player's relationship with a game's rules, and I was struck by the thought that, when both the rules and the actions of an opponent are determined/deterministic, a game is essentially static. Consider a computerized game of Tic-Tac-Toe with a simple AI. Once you figure out what the AI will do, and once you've mastered the rules, winning becomes a routine matter. You've "solved" the game, and it's only a puzzle that's trivial to complete. Extrapolating from this, other games present the same situation with more complex AIs rules. All games, therefore, are puzzles!

I thought I had come upon something profound, but now that I've more thoroughly examined my theory, I have to qualify it. And by "qualify it", I pretty much mean that I was wrong.

Costikyan introduces the first objection by observing that a game like Zork, which allows the player to travel freely through its world but otherwise consists entirely of puzzle-solving, is "90% puzzle and 10% game". The mere fact that you can proceed in an undefined path suggests that there's some adaptation to the player's actions. Still, this isn't entirely convincing; one might liken it to the ability of a player to walk away from a jigsaw puzzle for a while, or even to concentrate on a different piece of it. Just because I can control the presentation of a puzzle doesn't make it any less puzzle-like.

A more intuitive and damning objection to my generality is that the random elements that are sometimes present in games destroy the stasis that is the hallmark of puzzles. In less pretentious terms, puzzles can't be random, while games can. There are a couple of rebuttals to this, aside from a sticky argument involving physical determinism. With each random step in a game, the player is presented with a new situation that uses preset rules. To me, that just sounds like you're dividing the game into further puzzles. The game can still be "solved", but rather than taking a single step to the solution, the player must navigate a new puzzle after each random event. It's possible to have randomized games in which the player is incapable of winning (which would prevent "solving" the game), but as previously discussed, that just makes your game suck all the more.

What randomness is really trying to do, by introducing all of the game's sub-puzzles, is to increase complexity. If you know the rules by which the game operates (including the choices of opponents), then a game has been "solved". If we increase the complexity of the rules enough, then a game becomes impossible for a human mind to solve. Without the possibility of solving the game, players are forced to create their own heuristics and constantly estimate probability; in short, players are forced to make interesting choices. And that's when the fun comes in!

Tic-Tac-Toe isn't fun because it's too easy to solve; generally both players have solved it, too, so you know both your correct move and what your opponent will do. Checkers is a game that is just barely simple enough for AI to solve; human Checkers masters generally are too good to produce surprising games, but novices who haven't yet mastered the game can still have fun. Chess is complex enough that nobody has solved it, and it produces an engaging experience for players of all skill levels.

One simple way to increase the complexity of a game, aside from adding extremely complicated rules, is to support multiplayer. Humans are tough to predict, and so interesting choices abound. AI, in attempting to mimic the complexity of humans, has to resort to randomized elements and/or monstrously complicated algorithms that no human mind could discern. Dice Wars is an example of a game in which the AI is just barely simple enough to solve; the fun and frustration in the game come primarily from managing the various randomized situations that can arise.

So the takeaway for designers is this: don't let your game turn into a puzzle. Whether through multiplayer, randomness, or innate complexity, don't let the gameplay become so simple that it can be solved. The "correct response" should not be obvious at any level of gameplay. To ignore this is to risk losing all interesting choices in the game, which is, in turn, to risk losing all of a game's fun.