Rules are most of what defines a game, but there tends to be a method, a principle, by which a game's rules are applied, and these principles, the meta-rules that dictate the design of the game to the designers, are rarely evaluated. We tend to think that just because we make the rules that we have infinite control over the nature of the game world, but there's a step missing in that assumption: Just because we made the rules doesn't mean we necessarily made them self-consistent– and, even if we have, we haven't necessarily communicated them well to the player.
in other words, it's easy to make them broken from the start. If you make a broken watch, you haven't made a watch, you've made a locket full of gears– very handsome, I'm sure, but not functional at all.
I should clarify here, since I'm using the three terms 'rule', 'principle', and 'contract'. In this case, the distinction is between what the player is able to do in game space (rule), what the designer decides he is allowed to design (principle), and what the player perceives the game is allowed to do in the game space (contract). For instance, there's nothing in the rule set of a game like Super Mario World to dictate that there can't be a blind drop, which the player is forced to fall into, randomly arrayed with spikes and platforms. All of those are components of the game which are used in very similar configurations, so it's demonstrably within the rule set. However, the designers decided that it would be a principle of the game that it wouldn't contain trial-and-error challenges like that, and over time the game establishes an implicit contract with the player that it won't do things like that because they aren't 'fair'.
Creating the game's rules is a different task than creating the game's principles, and creating the game's principles is a different task than communicating these principles to the players as contracts. This can cause problems. Many games violate the contracts they seem to be conveying to the player, and this can be frustrating– after all, how are they supposed to know what to expect if you can't even keep the rules straight?
Here's an example: In one of the Suikoden games (a Japanese RPG series), there is a character who you can only recruit if you let his text scroll at normal speed rather than pressing a button to make it move faster. Now, pressing a button to make the text move faster is a convention of the genre, and it's never been implied in any game within that genre, save under this one circumstance as far as I am aware, that advancing the text means that your character is waiting impatiently, looking at his watch, tapping his fingers, and sighing. There's no implication that advancing text has any effect on the game world at all.
Sure, it's unexpected, it's a gameplay twist, but the level of unexpected we're dealing with here is along the lines of saying "I'm seriously not lying! I swear to god!" when, actually, you are totally lying. Yeah, it's kind of surprising, but only because no one expects anyone to be that much of an asshole. It's understood that certain user interface conventions are there for convenience's sake, and horrible form to punish the player for embracing them in that spirit– at least, doing so without some indication that you're doing something different with this game.
Not all players perceive a given game as having the same contracts. I've seen some people frustrated by puzzles and ideas which I loved in games because it violated the contract they expect of that game. For instance, there's one puzzle (a few actually) in Braid which requires the player to die to solve, and one person I know was infuriated at this, considering it, I think, 'ungamelike.' It was one of my favorite puzzles in the game, particularly since the solution occurred to me out of the blue when I was doing something unrelated, always a very satisfying way to solve a puzzle. This solution was absolutely consistent with the rules, and consistent with the principles of the game as the designer understood them– the design principles of Braid actually lie unusually close to the ruleset, one of its distinguishing characteristics– but because this is a behavior uncharacteristic of games, or at least the genre of games this player had associated the game with, the player felt betrayed.
Also worth noting is how well the contracts which seem to be most popular coincide with good storytelling. Just as a player encountering an unmarked trap and dying instantly feels cheap and unsatisfying, someone dying unexpectedly with no foreshadowing and for no apparent reason in a movie feels like lazy writing most of the time. I'm sure it's very realistic, very depressingly realistic, to have characters dying all the time for no good reason, but that doesn't make it good writing. No, in games we want to have the possibility of that stereotypical black lady yelling at us, "don't go in there, you gonna die! Aw, told you! Dude with a knife!"
(Which, now that I think about it, would be an excellent variation on the narrative formula used in Bastion. How is this not a thing? It would be amazing.)
Now, I'm not saying that you have to follow the established contractual conventions of your genre, but you do have to do your best to be aware of them. I don't think it's a coincidence that Valve, who established the principle of never taking control away from the player so boldly in Half-Life, was also the developer to break that principle more boldly and intelligently than any other developer in Left 4 Dead. For another interesting example of very intentionally breaking with the contractual conventions of a genre, look at I Wanna Be The Guy: hardly a perfect game, it must be admitted, but it is actually very smartly designed around the kind of sadistic traps it loves, the exact kind which Super Mario World scrupulously avoids, by making check points frequent and trial-and-error a hilarious-if-somewhat-infuriating experiment.
So here's a question to leave you with, one I've been pondering: Would it be possible to lie about the principles of the game, build false contracts up in the players' minds, and make that part of the game? To base the principles of the game on misleading your players and making them think it's a different kind of game than it is? Is it possible to invoke any emotions except resentment and betrayal using such a method?
Or could this be the unreliable narrator of the game design space? Could this be something new?