[OC] The same 14 strategies in Axelrod’s 1980 prisoner’s dilemma tournament, ranked by matches won and by points per round

—
SimulatedEcology

7 comments

  1. Tool: Python, NumPy and matplotlib; the tournament itself is my own implementation.
    Data: the fourteen strategies submitted to Robert Axelrod’s 1980 prisoner’s dilemma tournament, reimplemented from their published descriptions and played round-robin, 200 rounds per pairing. Every number in the chart comes out of that simulation rather than from Axelrod’s published table, so the ordering differs slightly from the original in the mid-table.

    A match is a “win” if one strategy finishes the 200 rounds with more points than the other. Downing wins all thirteen of its matches by small margins and gets destroyed on aggregate; Tit-for-Tat can never win by construction, since it only ever copies the last move.

    Video version, if anyone wants the walkthrough: [https://youtu.be/Ck1n8Z3W0lA](https://youtu.be/Ck1n8Z3W0lA)

  2. I don’t understand the subject matter here at all.

  3. I doubt that many people have the 14 1980 prisoner’s dilemma strategies memorized. IMO, the graph would be a lot of useful if it brief description of what the strategy is (and if that doesn’t fit with the names then just drop the names).

  4. Didn’t Tit for Tat win the original Axelrod’s tournament?

  5. Overview of strategies for those curious: https://mas.kke.co.jp/en/model/prisoner/

    Downing is basically “guess what they’ll do conditioned on what I did last time”

    Question for OP: what is the difference between matches won and points per round? I thought the actor with the most points won each round?

  7. Iterated prisoner’s dilemma is just an entirely different problem from a single instance.

Comments are closed.