Skip to contents

Runs n_games self-play games between the two policies with reproducible per-game sub-seeds.

Usage

mxo_simulate(
  policy_x,
  policy_o,
  n_games = 100L,
  config = mxo_config_default(),
  seed = NULL,
  record_eval = TRUE,
  progress = TRUE
)

Arguments

policy_x, policy_o

mxo_policy objects.

n_games

Integer scalar, number of games to play.

config

A config list (see mxo_config_default()).

seed

Optional integer base seed.

record_eval

Logical, passed through to mxo_self_play().

progress

Logical. If TRUE (default), display a cli progress bar.

Value

An object of class mxo_sim_result with components games (one-row-per-game tibble), records (list of mxo_game_records), policy_x, policy_o, config, seed.

Examples

set.seed(1)
sim <- mxo_simulate(mxo_policy("random"), mxo_policy("random"),
                    n_games = 3L,
                    config = mxo_config_default(n = 3L, ply_cap = 6L),
                    seed = 1L, record_eval = FALSE, progress = FALSE)
sim$games
#> # A tibble: 3 × 10
#>   game_id       seed winner outcome n_plies n_timelines win_axis_class
#>     <int>      <int>  <int> <chr>     <int>       <int> <chr>         
#> 1       1 1140350788     NA draw          6           4 NA            
#> 2       2  312928385     NA draw          6           3 NA            
#> 3       3  866248189     NA draw          6           4 NA            
#> # ℹ 3 more variables: cross_timeline_win <lgl>, first_move_kind <chr>,
#> #   first_move_idx <int>