Since Caylus is a perfect information game with absolutely no random elements, wouldn't it be possible to "solve" 2-player Caylus with a supercomputer? For each possible move of your opponent, it would calculate the tree etc. Or if it's as complex as go... then it's impossible

I think there are to many possibilities to (reasonably) build a tree of them.
First there are the different start setups of the pink buildings and who is start-player. 6 Buildings - 2 Players... means you have already 1440 different Start-Setups in a 2playergame.
As a Comparism: Chess and Go have only 2 Statsetups: I start or opponent start.
Then there'll be the first move.
Your oponent could spend 0-6 people on >10 possible fields.
So lets say 4 people on 10 different fiels for example that means 10*9*8*7=5040 different possibilities

Multiplied with the 1440 different setups you'll be at roughly 7 Mio possibilities after just one turn. Simple Math
Not having regarded yet, that he could have build different buildings and won grants by the king and a different number of buildings could be activated that round.

Sure - there will be much less different outcomes (but with the combination of different buildings in which order on plan, grants taken, winning points, goods stored etc. still a lot) and you COULD eliminate some possibilites because they seem not useful but I think my example shows, that building a tree of possibilities with such a complex game is not really an option.
You'll need to programm some kind of mechanisem to analyse the game situation - an Intelligence...

Termi

In this simplified example, you'll have to go with combinations instead of permutations, i.e., (10*9*8*7)/(4*3*2)=210

Can we have a Game Theory Fan geekbadge? Pretty please?


Anyway, maybe Caylus is solvable (the more i think about it the probable that is). However, who is gonna program this? You cannot go with raw computer power alone without some guidelines, opening theory and such. Have these things been established for Caylus in a universally accepted way?

Unfortunately i don't think IBM (or Microsoft) will be interested enough to put several M$ on this project

720. For purposes of solution it doesn't matter who goes first, only what they do.

However, you're right. The supercomputer that can solve chess doesn't exist, and that only has 400 possible game states after each player has moved once. Even assuming nobody plays to blank spaces or the Gold Mine on turn 1, in Caylus there are 151200 possible game states after each player has placed only a single worker. That's a tree 378 times larger than that of chess, and you haven't even come close to finishing the first turn.

I think you miscalculated. With two players, there can hardly be more than 100 to 200 states (too lazy to haul out the board and calculate exactly) because each player can only chose approximately 10 to 12 different locations on which to place a worker. Admittedly, if you increase the number of players the amount goes up, but the OP mentioned the 2P-game specifically.

There are 6! (720) possibilities in the setup alone, regardless of where people play. Then there are 15 places for the first person to play (6 pre-bridge, 8 after, and pass) and 14 for the second. Total: 151200. That is not considering the difference between the values of playing for the second person if the first person passed.

The different number of turns taken by each player and the change in values of placing workers after the other player passed would make this a much more challenging problem.

So you're too lazy to be bothered working it out or even counting the potential playing locations, but you still think you can turn around and criticise my numbers. Right. Your opinion has been noted and will be treated with the respect it deserves.

As it happens, I did miscalculate slightly - I forgot that there are two moves which can be made by any number of players. If the first player passes or plays in the castle, the second player has 15 options to choose from instead of 14. Adjusting, the actual number of possible game states after setup and first placements is 152680.

Not impossible!

In fact, let's challenge some computer geek to build a distributed computing solution. We can all run "the Caylus project" and given enough of us, we'll solve this thing in no time

Yesss!!!

Know any good computer/math geeks?

Something else we should work on, if it doesn't already exist, is a standard notation system to log games.

Any 2pl game with no luck is theoretically solvable; it's just a question of time. Thus, games like Chess, Go, and Caylus are all solvable ... but unlikely to be 'solved' any time soon.

By comparison, Tic-Tac-Toe is certainly solved, as is Connect Four.

What really intrigues me here is the idea of player 1 passing as their first action. Has anyone actually tried this in a game?