Skip to contents

A self-contained chess toolkit in native R: legal move generation, FEN/PGN I/O, position evaluation and alpha-beta search, batch simulation, an opening explorer, a puzzle trainer, and an interactive Shiny study app. An optional UCI hook (e.g. Stockfish) enables deeper analysis.

Why gambitR?

gambitR combines native R move generation with play, simulation, an opening explorer, a puzzle trainer, and an interactive study app. Move generation does not require an external chess engine; UCI analysis is optional.

Installation

# install.packages("remotes")
remotes::install_github("r-heller/gambitR")

Quick start

library(gambitR)

pos <- gm_position()
gm_legal_moves(pos)
gm_plot_board(pos)

# Read a game and replay it
game <- gm_read_pgn("1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 *", text = TRUE)
gm_identify_opening(game)

# Evaluate and search
gm_evaluate(pos)
gm_bestmove(pos, depth = 2)

# Play the native engine against a random mover, then batch it
gm_play(white = "engine", black = "random")
gm_simulate(white = "random", black = "random", n = 20)

# Train on the bundled puzzles
puzzle <- gm_puzzle(seed = 1)
gm_check_solution(puzzle, puzzle$solution)

# Launch the interactive app
gm_run_app()

Correctness

Move generation is verified against the published perft reference counts, with regression checks for castling rights, en passant, pinned pieces and promotion. These reference positions do not constitute an exhaustive proof:

Position Depth 1 Depth 2 Depth 3 Depth 4
Start position 20 400 8,902 197,281
Kiwipete 48 2,039 97,862
Position 3 14 191 2,812 43,238
Position 4 6 264 9,467 422,333
Position 5 44 1,486 62,379

These run as part of the test suite; depths 1 and 2 run everywhere, deeper counts are skipped on CRAN for time.

Local maintenance candidate

See MAINTENANCE.md for the 30 October 2026 repairs, exact local installation path, verification results and remaining release limits.