Psychology vs strategy games

Question

Chess is primarily a game of strategy, not psychology. By this, I mean that there is no bluffing or second-guessing involved. It should theoretically be possible to write an optimal program, 'Program A', which cannot be beaten by any other program. Even if another team of programmers is given the program, they will not be able to tailor-make a 'Program B', which is able to exploit Program A's weakness.

In contrast, many gambling games involve a heavy psychology element. I am reasonably confident that there does not exist a poker program 'A' which won't lose (on average) to any other program, even if that program is told exactly what program A is. I think one could tailor-make a program 'B' to exploit program 'A'.

Using this metric - whether or not it is possible to create a program that is unbeatable even if it's strategy is given away - would Sushi Go classify as a psychology or strategy game? What about other games?

Answer 1

I have no clue what “Sushi Go” is, but I can give insight into this little question: the major difference between the 2 games you described (chess and poker) is an element of uncontrollable chance. Poker is random, it cannot be “perfectly solved” because it relies on randomness. Games like chess on the other hand, have clear cut rules and it’s obvious that playing one move is better than another. Given enough time to run through all possible states and strategies, any computer could solve any finite game. However, I would like to point out that just because a computer plays perfectly, that doesn’t mean it always wins. In this sense, perfect just means that the computer always makes the strongest possible move. The computer can still tie. Take for example checkers (a game that has been solved), if the player plays perfectly against the computer, he can tie the computer.

Answer 2

I think that if you're trying to determine whether a game is based on strategy rather than psychology/bluffing, the test should be whether you can write a deterministic program which cannot be beaten by any other. For example, it is easy to write an unbeatable (on average) rock-paper-scissors program (choose an option at random), but you would presumably put it in the second category.

For multi-player games there is also the issue (hinted at in a comment) that the other players can conspire to beat you, even if you play as well as possible; this is even true in something like 3-player chess. So maybe the test for a multiplayer game should be as follows. We have two designated players A and B. The other players follow some strategy which doesn't distinguish between A and B (I guess we'd want A and B to be in opposite positions so there is symmetry there). You have to write a deterministic program to play A. Then your evil adversary gets to write a deterministic program to play B, having seen your program. Will he beat you, and how convincingly?

By this definition, I imagine Sushi Go has some "psychological" element. This is because knowing whether you are likely to go for the things where you get points for having most of, say, (I've played the game a few times but can't remember the details) is useful information for him in terms of deciding to go for those things or not bother. And actually this element would be a lot stronger if you're sitting closer together (because he'd have information either about what cards you're likely to pass or what cards you want him to pass you, depending on position). I'd still expect there's a significant amount of "strategy" though - plenty of options are just bad ideas irrespective of opponents' strategies.