3 ms·

I do this but to the rules of the board game rather than the physical pieces (though these merge in the case of custom maps) I'll use stratego as an example as

I do this but to the rules of the board game rather than the physical pieces (though these merge in the case of custom maps) I'll use stratego as an example as it's one of my absolute favorites and I'm a total arm-chair general. The game runs far too slow by default, and also subsequently doesn't simulate the notion of a "tempo" in battle. Before the players take their turns, a dice is rolled which decides how many moves each player makes in their turn. (I believe the game should be "fair", so it's locked as being the same number of moves for both players that turn.). I'm still trying to figure out if this introduces a significant first player advantage or not, and how that might be mitigated if it does. This rule change often makes stratego games playable in 10 minutes and massively opens up tactical options. I also think that the default stratego map is too small, and has far too generic terrain. I've got some larger custom maps, which include even more interesting terrain. Still trying to balance this, as larger maps significantly slow the game down. I'm also experimenting with the idea of using multiple moves to "pool" attacks, allowing using multiple moves to let larger numbers of units attack a single unit. I think that a powerful "general" absolutely should be capable of being overwhelmed by many less powerful units. Speaking of this, changing the whole idea of a units "power-level" makes sense to me. Strong units should be "special forces" or "elite", while the lower rankings should have buffs (i.e. generals are now rated as 2 but buff the ratings of all units next to them by +1) Like I can go on and on forever about rule changes to turn stratego into basically the board game battle simulator of my dreams. And yet I go to other peoples houses and see them stop playing their board games because of boneheaded default rules which cause the games to play too slowly.

vundercind 3y ago
Huh. If someone had asked me whether Stratego is even trying to be any amount of war simulation, even a bad one, I’d have said “no”. I’d class it as the kind of game where the flavor is just there to look at, and maybe as a mnemonic tool, and has little meaningful connection to the mechanics. Basically chess without all the information being public.
- Der_Einzige 3y ago
  Because of the limited information, I claim that Stratego would be much harder for machines to be good at. Still likely in the realm of "computer will destroy humans" but more like a game which would lend itself to rely less on alpha-beta pruning and monte-carlo tree search and the like (which are still used heavily even in NN evaluation using engines). Also means opening books are useless and end-game tablebases far larger. Humans may also find it easier to find "pathological" cases for AI Stratego models. Too bad no one in the AI community really cares about Stratego so not many engines exist. I claim adding the changes and similar changes to what I describe in my OP creates a "rougelikeification" of the game, starting to add elements of emergence and enough complexity to keep me entertained even in a world where increasingly large amounts of variations of all things become available to us via AI. I imagine that Stratego would be difficult to "solve" in a way that chess simply isn't. Of course, simply scaling up the size of the chess board and the number of pieces would have the same impact in chess...

vundercind 3y ago

Huh. If someone had asked me whether Stratego is even trying to be any amount of war simulation, even a bad one, I’d have said “no”. I’d class it as the kind of game where the flavor is just there to look at, and maybe as a mnemonic tool, and has little meaningful connection to the mechanics. Basically chess without all the information being public.

Der_Einzige 3y ago
Because of the limited information, I claim that Stratego would be much harder for machines to be good at. Still likely in the realm of "computer will destroy humans" but more like a game which would lend itself to rely less on alpha-beta pruning and monte-carlo tree search and the like (which are still used heavily even in NN evaluation using engines). Also means opening books are useless and end-game tablebases far larger. Humans may also find it easier to find "pathological" cases for AI Stratego models. Too bad no one in the AI community really cares about Stratego so not many engines exist. I claim adding the changes and similar changes to what I describe in my OP creates a "rougelikeification" of the game, starting to add elements of emergence and enough complexity to keep me entertained even in a world where increasingly large amounts of variations of all things become available to us via AI. I imagine that Stratego would be difficult to "solve" in a way that chess simply isn't. Of course, simply scaling up the size of the chess board and the number of pieces would have the same impact in chess...