I'd like to lay out a little mental model I find very helpful for game design. This model applies mostly to the mechanics and dynamics layers of a game, so I like to think of it in the aesthetic terms of a hedge maze.
A good game and a good hedge maze provide three things:
A space big enough that it's not immediately globally comprehensible (I can do many things, enough that I don't know everything I can do right away)
Constraints to keep the player's state locally comprehensible (Wherever I am, I have clear choices of what I can do)
Tools for navigating the global space from any given local space (I know where I am in relation to other places and I have the landmarks/signposts to get there)
In a hedge maze, the space, constraints, and navigation tools are all physical in nature. But in a game, they can be physical or abstract. A game space is the set of all possible states the game can be in, the constraints are the rules of the game, and the navigation tools are all the ways games help you know how to apply the rules to move between states.
As a maze designer, your minimum criterion is that players eventually reach the "goal state" (the exit). However, the fun lies in navigating to that state, and the steady drip of discovery along the way. In a small space, there's too little to discover. In an unconstrained space, navigation is trivial and discovery unsatisfying. In an unmarked space, everywhere feels the same and players get lost.
Consider the following mazes:
This maze is small and unconstrained with no landmarks. No fun to explore.
This maze is big enough to wander in, but its lack of constraints and landmarks make that wandering meaningless. Not much fun to explore.
This maze is constrained enough to provide clear choices, but its lack of scale and landmarks makes those choices meaningless. Not much fun to explore.
This maze has landmarks to navigate with, but its lack of scale and constraints makes that navigation meaningless. Not much fun to explore.
This maze has scale and constraints, but its lack of landmarks makes it unnavigable. Kind of fun to explore.
This maze has constraints and landmarks, but its lack of scale makes it burn out quickly. Kind of fun to explore.
This maze has landmarks and scale, but its lack of constraints makes it aimless and possibly paralyzing. Kind of fun to explore.
This maze has everything—scale, constraints, and landmarks. It has plenty to explore, constraints that provide clear choices, and landmarks that keep players oriented. Super fun to explore!
...and if we look at a real-world example like Longleat Hedge Maze, it matches the final example.
I've probably made my point by now for physical mazes, so let's look at how this applies to a more abstract space and use it to break down a game. I'll use chess, since most people already know how to play.
Here's how chess maps on to the parts of this model:
The game space is every possible arrangement of pieces on the board. The "exit" is checkmate, and the game is the process of navigating to that state.
The constraints are the rules by which pieces move and capture. For example, a pawn is constrained to moving forward one space at a time and capturing on a diagonal space.
The landmarks are simple observable facts that help you determine how close players are to checkmate, like whether the queen has been captured or a wall of pawns in front of the king.
Notice that although the board itself is a physical space, the game space, exit state, constraints, and landmarks are not physical in the same way that walls and towers in a maze are. Let's see what happens if we change some parts of the game.
One way to shrink the game space is by simply making the board smaller. If you play chess on a 4x4 board, it will get boring after two or three games because there just isn't very much game to play. Thus, it will be like navigating this maze:
One way to remove constraints is to let any piece move any number of spaces in any direction. This allows players to more freely navigate the game space, but it also makes decisions less meaningful and victory much simpler. Thus, it will be like navigating this maze:
One way to remove landmarks is to randomly select a piece on each team and make that the king, but tell neither player which piece it is. This makes it almost impossible to tell where players are in relation to the exit state and decisions become almost arbitrary. Thus, it will be like navigating this maze:
Thankfully, whoever designed chess put in all the features of this model and so made a pretty fun game. Hopefully this way of thinking helps you do the same!