Prisoner's dilemma (PD) and its iterated variant (IPD) is a well-known problem in game theory. The goal of this project is to create and compare optimization algorithms that find a good player for a particular IPD problem instance.
An IPD problem instance is given by
The payoff matrix must be set up such that the game is fair for both players. For 4 positive numbers a, b, c, and d, the matrix has the following form:
| P2 plays | |||
| 0 | 1 | ||
| P1 plays | 0 | (A,A) | (B,C) |
| 1 | (C,B) | (D,D) | |
where in each pair (r1, r2) denotes the rewards for player 1 and player 2, respectively.
Depending on relations among A, B, C, and D, the payoff matrix may be set up in such a way that
You should choose several interesting PM setups and try to find players for them.
There are many possible representations of the players/strategies:
The chosen representation shall be consulted with the teacher.
IPD is a two-player game. To evaluate the quality of a player is thus possible only by playing games against other players, i.e., the quality of a player depends on the set of chosen opponents. The choice of this reference set is up to you, it may be
The goal is to *maximize the sum of rewards across all played IPD games*.
In no particular order:
Take these questions as an inspiration of what you can study in your project.