Engines
Programs that play or solve games, each with the evidence for its strength and a link to it. A solver plays from a solution; an engine searches. Records are in data/engines/; npm run check rejects a strength claim without a source.
Strength evidence, strongest first: perfect-play, competition, rating, match-result, third-party, self-reported, none. A self-reported number is the author’s own and has not been reproduced here.
For games in this list (7)
Section titled “For games in this list (7)”Connect 4 Game Solver
Section titled “Connect 4 Game Solver”- Games: Connect Four
- Kind: solver · License: AGPL-3.0 · Author: Pascal Pons
- Strength (perfect-play): A solver for any Connect Four position; the companion tutorial, “Solving Connect 4: how to build a perfect AI”, builds it step by step. (source)
- Repository: https://github.com/PascalPons/connect4 · last checked 2026-10-17
- Notes: A web front end runs at https://connect4.gamesolver.org/.
dobutsu (Clausecker)
Section titled “dobutsu (Clausecker)”- Games: Dōbutsu shōgi
- Kind: engine · License: BSD-2-Clause (two-clause BSD text in the README; GitHub reports no SPDX match) · Author: Robert Clausecker
- Strength (perfect-play): “It uses a comprehensive endgame table base to achieve perfect play from any position.” (source)
- Repository: https://github.com/clausecker/dobutsu · last checked 2026-10-17
- Games: Othello
- Kind: engine · License: GPL-3.0 · Author: Richard Delorme
- Strength (third-party): Takizawa (2023), which used a modified Edax to solve the 36-empty-square positions in the weak solution of Othello, describes Edax as “among the strongest software under typical match rules”. (source)
- Repository: https://github.com/abulmo/edax-reversi · last checked 2026-10-17
- Notes: The modified Edax used for the solution is linked from the paper’s data-availability section.
- Games: Nine Men’s Morris, Lasker Morris, Morabaraba
- Kind: solver · License: GPL-3.0 (stated in the repository’s “License (gpl-3.0).txt”) · Author: Gábor E. Gévay & Gábor Danner
- Strength (perfect-play): Plays nine men’s morris, Lasker morris and Morabaraba perfectly from the authors’ strong-solution databases, and can compute those databases. (source)
- Repository: https://github.com/ggevay/malom · last checked 2026-10-17
- Notes: Windows only (Visual Studio solution). The released build is on the authors’ page, https://www.inf.u-szeged.hu/~danner/mills/.
MoHex 2.0 and Wolve (Benzene)
Section titled “MoHex 2.0 and Wolve (Benzene)”- Games: Hex (7×7), Hex (9×9, all openings), Hex (10×10, remaining openings)
- Kind: engine · License: LGPL-3.0 · Author: University of Alberta Hex group (Benzene); this CMake repository maintained by cgao3
- Strength (competition): Hayward’s Hex page: “Versions of MoHex have won gold respectively at every Computer Games Olympiad since 2009” (gold listed for 2009, 2010, 2011, 2013 and 2015). (source)
- Repository: https://github.com/cgao3/benzene-vanilla-cmake · last checked 2026-10-17
- Notes: This repository is a CMake port of Benzene with development to 2015, including the parallel DFPN solver behind the 9×9 and 10×10 f5 results (Pawlewicz & Hayward 2013). The Olympiad results are for 11×11 and 13×13 play, not the solved board sizes.
pentago (Irving)
Section titled “pentago (Irving)”- Games: Pentago
- Kind: solver · License: BSD (stated in the README; variant not named) · Author: Geoffrey Irving
- Strength (perfect-play): A strong solution: perfect play can be computed from any position, from a 3.7 TB database for 17 or fewer stones and by search from scratch above that. (source)
- Repository: https://github.com/girving/pentago · last checked 2026-10-17
- Notes: The solution can be browsed at https://perfect-pentago.net.
- Games: Gomoku (free-style), Renju (連珠)
- Kind: engine · License: GPL-3.0 · Author: Haobin Duan and contributors (AUTHORS file)
- Strength (competition): Won the freestyle 15×15, freestyle 20×20 and renju leagues at Gomocup 2025. (source)
- Repository: https://github.com/dhbloo/rapfi · last checked 2026-10-17
For other games (5)
Section titled “For other games (5)”- Games: Fog-of-war chess (dark chess)
- Kind: engine · License: GPL-3.0 · Author: Brian Liou
- Strength (none): No rating or match result published. The README says strength is gauged against humans on mistboard.com, and engine-vs-engine results only gate development. (source)
- Repository: https://github.com/brianhliou/misty · last checked 2026-10-17
- Notes: Maintained by this list’s maintainer.
MistyBanqi
Section titled “MistyBanqi”- Games: Banqi (Chinese dark chess)
- Kind: engine · License: MIT · Author: Brian Liou
- Strength (self-reported): Tuned by paired-deal self-play bakeoffs. The README makes no state-of-the-art claim: it has not played CLAP_CDC, DarkKnight or other top programs, so there is no head-to-head result. (source)
- Repository: https://github.com/brianhliou/misty-banqi · last checked 2026-10-17
- Notes: Maintained by this list’s maintainer.
MistyFlipJungle
Section titled “MistyFlipJungle”- Games: Flip Jungle (hidden-identity animal chess, 4×4)
- Kind: engine · License: MIT · Author: Brian Liou
- Strength (self-reported): Matches its own exact tablebase 99 to 100% of the time on endgames of up to five pieces, and agreed with a forward solver on all 200 midgame test positions; opening play is unverified. (source)
- Repository: https://github.com/brianhliou/misty-flip-jungle · last checked 2026-10-17
- Notes: Maintained by this list’s maintainer.
MistyJungle
Section titled “MistyJungle”- Games: Dou Shou Qi (Jungle)
- Kind: engine · License: MIT · Author: Brian Liou
- Strength (self-reported): W11-L2-D187 over 200 games (about +16 Elo) against the strongest open-source Dou Shou Qi engine the author found and ran; four-piece endgame play matches exact retrograde tablebases. (source)
- Repository: https://github.com/brianhliou/misty-jungle · last checked 2026-10-17
- Notes: Maintained by this list’s maintainer. The opponent engine is not named in the README.
PikaJieQi (Mistboard build)
Section titled “PikaJieQi (Mistboard build)”- Games: Jieqi (xiangqi with face-down pieces)
- Kind: engine · License: GPL-3.0-or-later · Author: The Pikafish authors (upstream jieqi_old branch); three commits by Brian Liou
- Strength (self-reported): “A strong player beats it.” Classical hand-written evaluation; no jieqi network with weights exists upstream. (source)
- Repository: https://github.com/brianhliou/pikafish-jieqi-wasm · last checked 2026-10-17
- Notes: A pinned snapshot of Pikafish’s jieqi_old branch with two engine fixes and a WebAssembly build; maintained by this list’s maintainer.