Seize the Throne is a board game in which the players control noble houses vying for power. Each player's house is comprised of several random nobles drawn at the beginning of the game.
All actions in the game -- from convincing minor houses to back your cause, to marching armies around, to engaging in courtly intrigues -- must be taken through having a particular noble in a particular place on the board.
Each noble has an epithet (i.e., "Aurelia the Pious"), and a special ability that matches their epithet. Nobles have a type - either Strong, Ordinary, or Weak - and each player begins with one noble of each type.
"Whichever player has most recently worn purple starts the game."
The notion of giving every noble a unique epithet ("Artemesa the Fierce", "Pennington the Debauched") provided far more thematic flavor than I expected. People asked to play the game again just to find out what new nobles would be available.
Tightly coupling all game actions to the physical noble pieces on the board has great potential as a game mechanic. Playtesters visibly fretted over whether to march their strongest nobles across the board to make risky moves or to hold them back for future opportunities.
I never found a way to make the existence of Weak nobles trigger genuinely difficult strategic decisions. Players generally found easy ways to keep cowardly nobles away from battle and boorish ones away from court.
I also never found a satisfactory victory condition. Using minor house support as a victory condition turned the game into a simple matter of arithmetic. Using seizing the royal court as a victory condition placed too mch emphasis on the (deliberately de-emphasized) military aspect of the game.
Physical board game design is far more like interaction design than I originally thought. Topics to consider include the list of core actions, the important data objects/playing pieces, and the various methods of displaying, sorting, and filtering them. Users and players both have very limited throughput in their visual cortex to absorb information and a very limited short-term memory in whch to store it.
The original draft of the rules was 15 pages long. I don't know what I was thinking either. If you're using more than a few game mechanics that players aren't already directly familiar with from other games, then no one is going to be able to figure out how to play your game.
Pay attention to what mechanics and sub-systems your players enjoy, because you might be able to repurpose and refine them into a different project.