i was listening to this podcast about cheating in chess and he mentions Laszlo Mero's model of class units for measuring game depth

i can't find the part about that in the book, and i'm not really sure how it's calculated :(

but it would be cool to use that kind of metric as part of a model to design a game

there could be rules constraining the types of game rules that could exist, like maybe it's a chess-like game with a fixed number of pieces with certain combinations of movement patterns

and then you could estimate the depth of a game by running simulations and measuring how much of an advantage a stronger computer has over a weaker computer

and use that to guide a search towards a game which has lots of interesting emergent strategy

chess is thousands of years old and went through a similar organic evolution process with different versions played in different cultures, and it's great, but i think we could probably invent a superior game

though, i'm thinking it would probably be a lot easier to begin in an area that is not chess, because it has a lot of complexity itself

a rudimentary minmax chess computer was one of the first things i ever tried to program as a kid and it was very hard! i never succeeded

maybe something like checkers where there's only one type of piece that does the same thing and few special rules

or even something more abstract like 'paper scissors rock' but adding more options or more hands or something

it would be nice for the optimisation process to have a way of detecting a 'dead game' that had trivial strategies to end in a forced win, but that is probably really hard to prove even for checkers