In implementing a physics based board game system, I don't need to program the
rules of the board game, "just" the rules of physics. Fortunately, the rules of physics are fairly consistent.
The same physics based system to play horse world, tic tac toe, or chess would work - its just about managing different models of objects in the world.
This implementation of chess ( https://youtu.be/XwHOH-C-4vA ) doesn't understand chess - its up to the human players to play it.
That it is a 3d physics simulator it allows humans to interact with it in the same way we would interact with the real world pieces.
There is no rules management / engine in the physics model preventing an invalid game move or determining when someone has won the game.
This is why its "easy" in that you just implement the physics engine, introduce the models into it and let the humans play the game.
If, however, you wanted to implement a game without the physics model, you would need a more complete model of the game state because you aren't using the 3d placement of the pieces and the humans to model it for you.
If you were to implement Ponte del Diavolo ( https://www.yucata.de/en/Rules/PonteDelDiavolo ) without physical pieces to place, how would you do that in a non physics system? Using that same game system could you implement Tally Ho! ( https://www.yucata.de/en/GameInfo/Halali ) or Sudoku Moyo ( https://www.yucata.de/en/GameInfo/SudokuMoyo )?
With table top simulator working off a physics model, ponte del diavolo is a board and two different types of pieces in two different colors. Tally Ho! is seven tiles (of various counts of each) and Sudoku Moyo is a board with nine sets of pieces with the numbers 1 through 9 on them (one in one color, and four sets in two different colors each).
The physics version of the game doesn't require a model of the game state, just the game pieces for two humans to play the game against each other.