Arimaa (pronounced uh-ree-muh) is a two-player board game invented by Omar Syed, a computer engineer trained in artificial intelligence. After Garry Kasparov was defeated by the chess computer Deep Blue, Syed wanted to create a game that could be played on a chessboard with chess pieces, but which would favor intelligence over calculating power. If a game had too many possible positions, a computer could not simply calculate its way to victory. Syed soon realized, however, that a large branching factor would not necessarily help humans, who would need to anticipate a position a few turns ahead. Syed began to see the solution while teaching his young son a simplified version of chess; if movement was limited but each turn allowed for multiple steps, a game could be both high-branching and playable. After much experimenting, Syed came up with a game which, like chess, used a 1-1-2-2-2-8 distribution of pieces. Conceptually, Syed's new game was much simpler than chess, yet the branching factor dwarfed that of chess. For a human or machine, imprecise planning would be key.
Arimaa's first computer test was the Zillions of Games engine, which was designed to competently play any game it was given the rules to. As Syed expected, Zillions was an easy opponent. On November 20, 2002, Syed published the rules for Arimaa and announced a $10,000 prize, available annually through 2020, for the first bot to defeat a top human player in a "Challenge" match. David Wu's bot Sharp claimed the prize in 2015.
United States Patent number 6,981,700 for Arimaa was filed on the 3rd of October 2003, and granted on the 3rd of January 2006. Omar Syed also holds a trademark on the name "Arimaa". Syed has released an experimental license called "The Arimaa Public License", with the declared intent to "make Arimaa as much of a public domain game as possible while still protecting its commercial usage". Items covered by the license are the patent and the trademark.
Arimaa can be played online at the arimaa.com gameroom. Elsewhere, it can be played using a chess set, since each Arimaa piece corresponds to a chess piece. In 2009, Z-Man Games began producing a commercial Arimaa set. Only one face-to-face tournament has taken place, but various events have been hosted online:
- The World Championship is an annual tournament for human players. This tournament has taken various formats, and has sometimes lasted several rounds. Jean Daligault of France has been the world champion six times, in 2007, 2009, 2010, 2011, 2013, and 2014. Mathew Brown prevailed in 2015, 2016, and 2017. Karl "Fritz" Juhnke won in 2005 and 2008.
- From 2004 to 2015, the Computer Championship matched the top Arimaa bots in an elimination tournament. David Fotland's program Bomb placed first from 2004 to 2008. Jeff Bacher's Clueless prevailed in 2009. Mattias Hultgren's Marwin won in 2010 and 2012. David Wu's Sharp triumphed in 2011, 2014, and 2015. Ricardo Barreira's Ziltoid was victorious in 2013.
- The Arimaa Challenge took place following the Computer Championship. At first, the winning bot played an eight-game match against a lone human defender. The format evolved: from 2007 onward, the top two bots were available to play during a "screening" period. A human could play two games against each of these two bots; the bot with the better record in those matches advanced to the Challenge, where it faced three ultimate human defenders, who were selected beforehand and hadn't played in the screening. If this bot could win best-of-three matches against all three human defenders, its developer would win the $10,000 prize. Humans dominated until 2015, when Sharp defeated Jean Daligault, Lev Ruchka, and Mathew Brown.
In addition to these events, the 1st Arimaa Online Festival was organized for 11 September 2010. This event included Arimaa matches, a strategy workshop, and an interview with Omar Syed.
After DeepMind's AlphaZero mastered Go, Chess, and Shogi simply by playing itself, Omar Syed announced a $10,000 prize for the creation of such an Arimaa bot which could win a 10-game match against Sharp. This has not yet been done, although a bot-in-progress named "Dolores" appears promising.