Quote:There's not really much else to it - if certain strategies cost more to choose, or cost more to switch between, that factors in, but there's still ultimately an optimal metastrategy which, if employed once against an infinite number of random players, will win more often than any other metastrategy. The only hard part is gathering the data on 'costs'.
If we are talking about random enemies as opponents, then for chess the clearly optimal strategy is the Fool's Mate as a random enemy will be very unlikely to counter it. There are thousands of potential strategies and only a handful counter the fool's mate, so clearly, using that strategy is the bet because the enemy is unlikely to use the counter strategy and it finishes the game quickly so it limits the chance that you opponent can mount an attack on you (you can't get a quicker mate than the fool's mate).
So under your own criteria for deciding what is an optimal strategy (or even meta strategy) Chess is an exceedingly shallow game as there is one very simple strategy that beast any set of random players.
However, what makes Chess so much better is the Human factor. The fact that they can bluff, adapt their strategies, spot simple tricks (like the fool's mate), etc.
Statistically, if we assume players have a complete write up of all possible strategies, in chess Black has a better chance to win as their opening moves tend to put White on the defensive. This has not been completely tested (as no one has calculated all possible outcomes of all possible chess games using all possible moves available), so it might just be that there are some yet to be discovered "games" of chess that evens this out. But as far as we know, black has the advantage in chess.
If we therefore eliminate the Human factor, then chess is a broken game. Black wins. If players only ever did the best possible move in chess, white would never win.
But, then why do people still play chess? Well it is because of the Human factors. Fallible and sneaky humans are in charge of playing the game, so they can do things that would not be "optimal" and by doing so, gain or loose the advantage.
Also, lets take Poker. This is a game that, if only the rules and their interactions are studied (that is assume that players only make perfect decisions), then it is only a game of chance.
But, why is it then, that there can be good Poker players and bad poker players? Again, it's the human element. Humans can bluff, make mistakes, take risks, get scared, etc. It is this human element that make poker a game worth playing, that k makes it a complex game.
People have tried to use game theory to work out the perfect poker strategy and it is quite trivial to do so. However, the player who work out the perfect strategy still loose to a good poker player. Why? Again, the human ability to bluff, make mistakes and take risks can throw the "perfect" or "optimal" strategies out and beat them.
This just shows that Game Theory does not always apply to Game Practical (that is real games played for fun or profit).
By considering the player and their psychology as part of the game system, this can massively increase the complexity of a game. This is why multiplayer games are so popular. The players like the added complexity and unpredictability (not just randomness) that playing real humans confers.
Quote:Most of the rest you talk about involves anticipating future actions and reactions, which is outside the scope of what I'm interested in, because although it is interesting it its own right, it really just ignores the issue by substituting temporal changes in choices for initial breadth of choice.
Most real games are repeated "games", therefore future actions and reactions are an important aspect of any strategy.
In an RTS, if you produce lots of Unit type A (which can be beaten by Unit type B), and the enemy knows this (which might be from a skirmish earlier), then they will build Unit type B. At the point when the second player learns of the first player's choice, the first player must consider the second player's future actions (and my choose to build Unit type C to beat the Unit type B that the second player is likely to produce).
Quote:I'm trying to avoid any metagaming aspects, especially iterative versions of a game (eg. iterative RPS, or iterative Prisoners' Dilemma), because I don't think they are relevant here
The fact that a game
is repeated adds in a lot of complexity (which is what this thread is about). So, if you are looking for an easy way to add in complexity to a game, here is the simplest way: Make future strategic decisions be based to some degree on information learned from an earlier run-through of the game. In other worlds, make the game repetitive.
Quote:interaction between multiple aspects seems to be the key here, as someone mentioned above, but making that interaction non-trivial is important too.
Looking at this concept, I think the best way to make the interactions non trivial is to make the actions of individual components (units, choices, etc) not be directly related to the strategy. This is what chess does.
In chess, the moves of each individual piece does not directly relate to the overall strategy. However, as the relative interaction of the pieces (their placement on the board relative to each other) determines the strategy.
So with a simplified RTS, you have a Scissor/Paper/Rock relationship between the units. As the strategy relates to what proportion you have between them, the individual pieces (the units in this case) directly relate to the strategy.
By creating a layer of indirection between the components that are used to create a strategy and the strategy it's self, this will make it harder for a player to create an optimal or perfect strategy as the same set of component could be parts of a different strategy.
Also, by having components indirectly relate to the strategy, it means that changing some of the pieces doesn't necessarily change your strategies. Like in chess, pieces that are not directly being used in an attack can still be moved without effecting that particular attack strategy (and may be part of separate or possibly complimentary strategy). The move might be positive, negative or completely neutral to the overall game, and still have no effect on the current strategy being used (and a piece can also be in several potential positions and have the same strategic effect).
If the pieces in chess were directly linked to the strategy, then a strategy would only be described by the positions of the pieces on the board and changing any piece would change the strategy.