| Docs Build |  |
| Documentation |  |
| GHA CI |  |
| Code Coverage |  |
| Bors enabled |  |
An experimental package for developing poker bots
Here, we'll battle 4 bots (two RandomBots with prob = 0.6, and two with prob = 0.5):
Click to expand and see the bot definitions
struct BotRandom <: AbstractAI
prob::Float64
end
function TH.player_option!(game::Game, player::Player{BotRandom}, ::CheckRaiseFold)
prob = TH.life_form(player).prob
if rand() < prob; check!(game, player)
else
amt = Int(round(rand()*bank_roll(player), digits=0))
amt = TH.bound_raise(game.table, player, amt) # to properly bound raise amount
raise_to!(game, player, amt)
end
end
function TH.player_option!(game::Game, player::Player{BotRandom}, ::CallRaiseFold)
prob = TH.life_form(player).prob
if rand() < 1-prob
if rand() < prob; call!(game, player)
else # re-raise
amt = Int(round(rand()*bank_roll(player), digits=0))
amt = TH.bound_raise(game.table, player, amt) # to properly bound raise amount
raise_to!(game, player, amt)
end
else
fold!(game, player)
end
end
function TH.player_option!(game::Game, player::Player{BotRandom}, ::CallAllInFold)
prob = TH.life_form(player).prob
if rand() < 1-prob
if rand() < prob; call!(game, player)
else; raise_all_in!(game, player) # re-raise
end
else
fold!(game, player)
end
end
function TH.player_option!(game::Game, player::Player{BotRandom}, ::CallFold)
prob = TH.life_form(player).prob
if rand() < 1-prob; call!(game, player)
else; fold!(game, player)
end
endOnce the bots have been defined, battles can be configured in this repo as follows (see test/runtests.jl for the full implementation):
n_tournaments = 50
bots = (
PB.BotRandom(0.4),
PB.BotRandom(0.4),
PB.BotRandom(0.6),
PB.BotRandom(0.6),
)
@time stats, n_games = PB.battle!(n_tournaments, bots...)
plot_stats(stats, n_games)Here's the results. At 50 tournaments it's (sometimes) difficult to see which strategy is doing better
At 500 tournaments, it's clear which strategy is winning out
Note that there is variance between the two exact same strategies-- that's the "luck" of the game! But it's clear that the exact same strategies have the same asymptotic behavior.
| Package | Development status | Purpose |
|---|---|---|
| PlayingCards.jl | Perhaps stable | Representing cards |
| PokerHandEvaluator.jl | Perhaps stable | Comparing any two 5-7 card hands |
| TexasHoldem.jl | Likely changes needed | Simulating multi-player games of TexasHoldem |
| PokerBots.jl | very early development | Battling bots with prescribed (or learned) strategies |
| PokerGUI.jl | very early development | Visualizing TexasHoldem games via a GUI |