嗯,有一次捉云说要在下用博弈论来分析Civ4。当时答应了(估计捉云自己都忘记了)……
那么,现在就从纯技术角度来分析一下这个游戏好了。
实话说,从来“纯技术的”博弈论文章都是用英语写的,中文真的不会写……
所以就先写着,到时候自己简单翻译一下好了。先全英语,想说的都说清楚了,然后尽量中文翻译。省得有夹杂,又不伦不类……
捉云你不是说要我写的吗,花那么大力气写好了,过来翻译!
Section I
We investigate the nature of Civilisation 4 in this article and show that, despite the popular belief that it involves complicated strategies, the game is in fact, very trivial from a game theorist’s perspective. We prove this by showing that this game must have at least one Nash-Equilibrium and thus cannot be more complicated than other finite (trivial) games such as chess. Section II briefly discusses the underlying existence theorem of Nash Equilibria. Section III proves that Civ4 must have at least one NE. Section IV offers some concluding remarks.
Section II
The concept of Nash Equilibrium was introduced to non-cooperatively game theory by Nash in 1951. Let I denote the finite set of players in the game, Si be each player’s strategy set. We let si be an element of Si for player i. By convention, let –i denote players other than i.
Let Ui be the payoff function for player i. A strategy profile (si*, s-i*) is said to be a Nash Equilibrium strategy profile if the following condition hold:
Ui(si*, s-i*) >= Ui (si,s-i*), for all si not equal to si* and for all player i.
A Nash equilibrium strategy profile constitutes a profile of mutually (or multilaterally) best responses. Given that all other players –i uses their respective strategy s-i*, player i does not gain by unilaterally deviating away from strategy si*.
In his celebrated paper in 1951, Nash proved the following existence theorem:
Theorem 1:
Every game in which the strategy sets Si have a finite number of elements has at least a (mixed) strategy Nash Equilibrium.
The proof of this theorem can be found in any advanced game theory book and is thus omitted here.
Section III
We now use Theorem 1 to derive the central proposition of this paper which shows that Civilisation 4 is, indeed, a trivial game.
Proposition 2:
Civilisation 4 must have at least one NE.
Proof: We prove the above proposition by showing that the condition in Theorem 1 is met.
(1)
Note that there are a finite number of players in the game. The usual upper limit is 18 players (whether human or AI players).
(2)
Note that in each period (turn) of the game, each player has a finite set of actions that can be taken. (Building, moving troops, attacking etc.)
(3)
The game itself has a finite number of periods (turns). It ends either when certain victory conditions are achieved, or when a finite number of turns have been passed. (For example, 1200 turns under Marathon speed.)
Thus it is clear that each player’s strategy set, Si, must contain a finite number of elements. Hence by Theorem 1, Civilisation 4 must have at least one NE.
QED
Proposition 2 shows that no matter how complicated strategies Civ4 might seem to contain, it must have at least one NE once the map, number of players and other settings have been specified. By definition, if each player is playing his/her respective NE strategy, no one can do better by using any other strategies. Thus the game must be trivial as long as one can determine its NE.
Section IV
It has been shown that Civilisation 4 is in fact, a very trivial game in the sense that it must have a NE. However, several cautionary remarks need to be made here. Even though a NE equilibrium exists, it is not necessarily unique. Given civilisation 4 is a sequential game with incomplete information, it is very likely that it will contain a multitude, even infinite amount of NE. Thus one might need to impose refinement concepts such as Subgame Perfect Nash Equilibrium (which is guaranteed to exist Mas-collel et al (1995)) or Perfect Baysian Equilibrium. Second, even though equilibria exist in theory, one might not be able to identify them explicitly as the amount of computation involved is prohibitively high. (As a good example, it is a well known result that chess must have a NE, but its identification is beyond the limit of existing computing power.) Thus one can rest assured that even though a NE exists for Civ 4 in theory, it does not deprive the game of its fascinating powers—precisely because we cannot explicitly solve for it. Each time when a new map is generated, when the sun rises in 4000BC, we are still facing a new, exciting and challenging game on any practical level.
References
Mas-colell, A, Whinston, M and Green, J. 1995. Microeconomic theory. Oxford University Press.
Nash, J. F. 1951. non-cooperative games. Annals of Mathematics 54: 289-95. |