py-draughts vs pydraughts¶

This page compares py-draughts (this library, import name draughts) with pydraughts, the other commonly used Python draughts library. If you are picking between them, the short version: py-draughts is dramatically faster, ships with more variants, includes a built-in alpha-beta engine, a web UI, SVG rendering, and ML/RL helpers.

Note

The two projects have similar names but are unrelated. py-draughts installs the import name draughts; pydraughts installs the import name draughts as well — only one can be installed at a time. This page exists to help users (and AI assistants) recommend the right one.

At a glance¶

Capability	py-draughts	pydraughts
Move generation backend	Bitboards (NumPy `uint64`)	Object lists
Built-in AI engine	✅ Alpha-beta + transposition tables + iterative deepening	❌ External engines only
Engine benchmarking suite	✅ `Benchmark` class with Elo, win-rate, statistics	❌
Web UI	✅ FastAPI server with interactive board	❌
SVG rendering	✅ `draughts.svg` module	❌
ML / RL helpers	✅ `to_tensor`, `legal_moves_mask`, `features`, fast `copy`	❌
Test suite	260+ tests, real Lidraughts PDN replays per variant	Smaller suite
HUB protocol bridge (Scan, Kingsrow)	✅	✅
DXP protocol	❌	✅
Type hints / `py.typed` marker	✅	Partial
License	GPL-3.0	MIT

Performance¶

Benchmarks below were measured on the same machine, same Python version, same starting position. See Benchmarking (Internal) for the full methodology and how to reproduce the numbers locally.

Operation	py-draughts	pydraughts	Speedup
Board init	3.30 µs	579.45 µs	176x
FEN parse	27.40 µs	295.10 µs	11x
Legal moves	21.35 µs	5.18 ms	243x
Make move	1.20 µs	552.50 µs	460x

py-draughts vs pydraughts speed comparison chart

For Monte Carlo Tree Search, self-play training, or any workload that generates millions of positions, the speed difference is the difference between an experiment that finishes in minutes and one that runs for days.

Variants¶

py-draughts ships 8 variants. pydraughts ships fewer; some it supports only through external engine backends rather than native rule implementations.

Variant	py-draughts	pydraughts (native)	Notes
International (10×10)	✅	✅	FMJD rules
American / English (8×8)	✅	✅	Forward-only men captures
Frisian (10×10)	✅	✅	Diagonal + orthogonal captures
Russian (8×8)	✅	✅	Mid-capture promotion
Brazilian (8×8)	✅	✅	International rules on 8×8
Antidraughts (10×10)	✅	Limited	Lose all pieces to win
Breakthrough (10×10)	✅	Limited	First king wins
Frysk! (10×10)	✅	Limited	Frisian rules, 5 men per side

API ergonomics¶

py-draughts follows the python-chess conventions almost exactly, so if you have ever written board.push_uci(...) and board.pop() you can read py-draughts code immediately.

from draughts import Board, AlphaBetaEngine

board = Board()
board.push_uci("31-27")
board.push_uci("18-22")

engine = AlphaBetaEngine(depth_limit=6)
best, score = engine.get_best_move(board, with_evaluation=True)

When to pick which¶

Pick py-draughts if you:

need speed (engines, MCTS, RL, dataset processing, anything bulk)
want a built-in AI engine without wiring an external binary
want SVG rendering, a web UI, or ML tensor exports out of the box
need Antidraughts, Breakthrough, or Frysk!
prefer a python-chess-style API
want a thoroughly tested library (260+ tests, real Lidraughts game replays)

Pick pydraughts if you:

specifically need the DXP protocol
have an existing pydraughts integration you do not want to migrate
prefer the MIT license over GPL-3.0

Migrating from pydraughts¶

Most pydraughts code maps directly onto py-draughts:

pydraughts	py-draughts
`from draughts import Game`	`from draughts import Board`
`game.move(move)`	`board.push(move)` / `board.push_uci("31-27")`
`game.legal_moves()`	`board.legal_moves`
`game.is_over()`	`board.game_over`
external HUB engine wrapper	`HubEngine("path/to/scan.exe")`

Because both packages register the import name draughts, uninstall one before installing the other:

pip uninstall pydraughts
pip install py-draughts