Game Theory with Applications to Finance and Marketing I, Lecture 3 & 4 Notes (Oct. 06, Oct. 13)
Game Theory with Applications to Finance and Marketing I, Lecture3 Notes(Oct. 06 & Oct. 13)
Math Review
Let’s review some math tools for Game Theory
.
Set and Function
Set
**Definition 1 (Set, Collection, Family): **
- A well-defined collection of objects (called elements) is a set;
- A set of sets is usually referred to as a collection;
- A set of collections is usually referred to as a family.
**Definition 2 (Empty Set): **A set that contains nothing is called an empty set, denoted by $\phi$.
Definition 3 (Super set and Sub set): **If two sets $A$ and $B$ are such that $B$ contains each and every element of $A$, then we say that $A$ is a **subset of $B$ and $B$ a superset of $A$, we write $A \subset B$.
Hint: If we want to judge whether set $A$ is equal to set $B$, we need: \(A \subset B, B \subset A\) *Definition 4 (Complement): **A set $A$ has its complement set, denoted by $A^c$, which is, \(A^c = \{x: x \in X, x \notin A \}\) where set $X$ is a *universal set.
**Definition 5 (Power set): ** Denote by $2^X$ the collection containing all subsets of $X$, which will be referred to as the power set of $X$.
Function
**Definition 6 (Function): **Given two sets $A$ and $B$, an assigning rule $f$ is a (single-valued) function from $A$ into $B$, if: $\forall a \in A$, there is $b \in B$, such that
$f(a) = b$. We can write: \(f: A \rightarrow B\) and 𝐴 and 𝐵 are called the domain and range of 𝑓, respectively.
Hint:
- If $A$ is a collection of sets, we say that $f$ is a set function.
**Definition 7 (Surjective, injective, and one-to-one): **
- Surjective (满射)
如果有$B \subset f(A)$,则我们称函数$f$是一个满射。
- Injective (单射)
如果满足: \(x, y \in A, x \neq y \Rightarrow f(x) \neq f(y)\) 我们称函数$f$是一个单射。
- Bijective (双射,又称一一对应, one-to-one correspondence)
如果一个函数又是surjective,也是injective,则称其为Bijective.
Image and Pre-image
**Definition 8 (Image): **
Nash Equilibrium存在定理
在求解均衡前,要判断是否存在Nash Equilibrium。因此,需要利用几个定理来判断。
Theorem 1 (Wilson, 1971; Harsanyi, 1973) : NE个数
Almost every finite strategic game has an odd number of NE’s in mixed strategies; more precisely, with the set of players and their strategy spaces fixed, the set of payoff functions that result in a strategic game having an even number of NE’s has zero Lebesgue measure.
- 每一个有限赛局都有奇数个混合策略的Nash均衡
- 即:混合策略NE+单纯策略NE=奇数
这个定理让我们明白了重要的道理。即,如果要找NE,则必须找出奇数个。如果只找到偶数个,就请再继续接着找。
Theorem 2 (Nash, 1950) : 混合策略存在定理
Every finite normal-form game has at least one mixed-strategy NE (which may or may not be a pure-strategy NE).
Theorem 2 is actually a special version of the following more general theorem (see Fudenberg and Tirole’s Game Theory, Theorem 1.2). Which is,
Nash的这个定理告诉我们,只要是有限的赛局必定有至少一个混合策略均衡。
Theorem 3 (Kakutani, 1941) : 不动点存在定理
这个定理是一个不动点存在定理。意思是:
一个定义在非空、凸集合、紧致集合的函数$r$,如果$r$上半连续,则存在一个不动点$\sigma^$,使得$r(\sigma^) = \sigma^*$成立。
显然,如果r是一个single value的函数,则r必然连续。
Function $r(·)$ in Game Theory :: What is “response function” ?
所谓的response function,是指player i在其他玩家采取策略时,自己采取什么策略。函数的input可以理解成是其他玩家的策略集合$\sigma_{-i}$,输出的是player i采取的策略$\sigma_i$。
Proof: Use Thm 3 to proof Thm 2
这部分略过。需要注意的是,NE的概念就是不动点的概念。用数学符号表达应该是:
- 假设$r(·)$是所谓的反应函数,$\sigma_i$是player i使用的混合策略。
- 本身而言,Nash均衡类似于:对方出招1,我出招2,对方必定跟着出招1。
- 假设$\sigma_$是所有player的策略集合所组成的向量。则有,如果response(反应函数)产生的反应策略结果还是$\sigma_$,则这个$\sigma_*$一定就是纳什均衡。
- 这就转化成了找不动点的问题。
没有Nash Equilibrium的情况?
要证明没有Nash Equilibrium,只要证明反应函数$r(·)$不存在不动点即可。
- 一个典型的例子是,非连续的payoff function可能导致NE不存在。
Theorem 4 : Mixed Strategy NE?
同时选择的赛局(simultaneously choose game),payoff function连续,则该赛局必定有一个混合策略的NE。
一个类似的定理(Theorem 4’)是:
对于一个有限赛局,若:
- 只有有限个player
- 策略集合是非空紧致集合(nonempty compact set)
- 对所有的player i,效用$u_i$是连续函数
则赛局有一个混合策略的NE。
Theorem 5 : NE in Symmetric game
A finite symmetric game has a symmetric Nash equilibrium in mixed strategy.
Reference
This note is mainly derived from Chyi-mei Chen’s Game Theory with Applications to Marketing and Finance, I
Course @ National Taiwan University. Thanks him for giving this kind of excellent lecture.