Definition

A strategy vector σ\sigma^* (in mixed strategies or behavior strategies) in an extensive form game Γ\Gamma is called a subgame perfect equilibrium if for every subgame, the restriction of the strategy vector σ\sigma^* to the subgame is a Nash equilibrium of that subgame: for every player iNi \in N, every strategy σi\sigma_i, and every subgame Γ(x)\Gamma(x),

ui(σ|x)ui(σi,σi|x)u_i(\sigma^* \vert x) \geq u_i(\sigma_i, \sigma_{-i}^* \vert x)

Remark

Since every game is a subgame of itself, by definition, every subgame perfect equilibrium is a Nash equilibrium.

Definition (subgame)

Ji(k)(μi,μi|hk)=t=kTui(t,μi(t,ht),μi(t,ht)|ht)J_i^{(k)}(\mu_i, \mu_{-i} \vert h_k) = \sum_{t=k}^T u_i (t, \mu_i(t, h_t), \mu_{-i}(t, h_t) \vert h_t) {μi(t,ht)}t=kT\{\mu_i(t,h_t)\}_{t=k}^T

SPNE for strategy profile in terms of horizon restriction

A strategy profile {μi}iN\{\mu_i^*\}^N_i is a subgame perfect Nash equilibrium (SPNE) if for any kk, and its associated hkHkh_k \in H_k, the restriction {μi}i=1N\{\mu_i\}^N_{i=1} to the horizon t{k,k+1,...,T}t \in \{k,k+1,...,T\} is also a Nash equilibrium.

Ji(k)(μi,μi|hk)Ji(k)(μi,μi|hk)J_i^{(k)}(\mu_i^*, \mu_{-i}^* | h_k) \geq J_i^{(k)}(\mu_i, \mu_{-i}^* | h_k) for every feasible μiΓi\mu_i \in \Gamma_i, $h_

Notes:

SPNE for infinite horizon game

#incomplete

Theorem

In an extensive form game without non-trivial subgames, every Nash equilibrium (in mixed strategies or behavior strategies) is a subgame perfect equilibrium.

Theorem

Let σ\sigma^* be a Nash equilibrium (in mixed strategies or behavior strategies) of an extensive-form game Γ\Gamma, and let Γ(x)\Gamma(x) be a subgame of Γ\Gamma. If 𝐏σ(x)>0\mathbf{P}_{\sigma^*}(x) > 0, then the strategy vector σ\sigma^* restricted the subgame Γ(x)\Gamma(x) is a Nash equilibrium (in mixed strategies or behavior strategies) of Γ(x)\Gamma(x).

See also


References

  1. M. Maschler, E. Solan, and Shmuel Zamir, Game Theory, Cambridge University Press, 2013, pp. 254-255.