Strategy simulation for simple games.
] add SquidGameUsing function play_game, simulates a game for of N strategies competing against each other (where N is the appropriate number of players for the game).
Implementable games are "deterministic" games where each player knows the possible rewards for each round at decision time, with the only uncertainty being the action of other players.
Rewards are defined as a multi-dimensional array where your action is the first index and rivals are the remaining indexes. For example, a 2 player game:
| Rival Action 1 | Rival Action 2 | |
|---|---|---|
| Your Action 1 | 5.0 | 0.0 |
| Your Action 2 | 10.0 | 2.0 |
In this case:
- If you acted with action
1and your rival also chose action1, you would both earn5points; - If you acted with action
1and your rival chose action2, you would earn0and they would earn10.
using SquidGame
import SquidGame._run_strategy
using Plots
number_of_rounds = 10
# Define Game
# (The infrastructure is generic enough so you can have different rewards per iteration).
prisoner_game = prisoner(number_of_rounds) # same as `Game(; rewards=(iter) -> [[5. 0]; [10 2]], number_of_rounds)`
# Define a name for your strategy
abstract type MyStrategy <: Strategy end
# Define the logic of your strategy
function SquidGame._run_strategy(::Type{MyStrategy}, reward::AbstractArray{Float64},
history::Union{NamedTuple{(:my_action, :their_action, :my_reward, :their_reward), Tuple{Vector{Int64}, Matrix{Int64}, Vector{Float64}, Matrix{Float64}}}, Missing},
rounds_left::Union{Int, Missing}
)
# your can use whatever logic you wish. This example is an angel, it will always choose the cooperative action.
return find_cooperative_action(reward)
end
strategies = Vector{Type{<:Strategy}}()
# push it to the stage
push!(strategies, MyStrategy)
# Push your opponent
push!(strategies, Devil)
# Simulate the game.
realized_reward_history, strategies_action_history = play_game(prisoner_game, strategies)
# Visualise the scores over all rounds
scoreboard(realized_reward_history, strategies)
# add a player
push!(strategies, RandomStrategy)
# choose the game
random_game = randgame(number_of_rounds)
# Simulate the game
realized_reward_history, strategies_action_history = play_game(random_game, strategies)
# Visualise the scores over all rounds
scoreboard(realized_reward_history, strategies)