Go: Difference between revisions
(Updated advancement of computer Go playing) |
(Fixed the links) |
||
Line 4: | Line 4: | ||
Go is far behind [[Chess]] in AI. While the highest level chess players can be defeated by AI, AI for Go struggles even with major handicaps. This is because when the game is played on a standard full sized board, there are '''far''' more possible games than in chess, and so even a supercomputer will struggle to find a good strategy through brute force and we have yet to come up with a more efficient AI that works well. So far, only less popular variants with a much smaller board have been solved, and it's possible that larger versions of the game will never be solved. | Go is far behind [[Chess]] in AI. While the highest level chess players can be defeated by AI, AI for Go struggles even with major handicaps. This is because when the game is played on a standard full sized board, there are '''far''' more possible games than in chess, and so even a supercomputer will struggle to find a good strategy through brute force and we have yet to come up with a more efficient AI that works well. So far, only less popular variants with a much smaller board have been solved, and it's possible that larger versions of the game will never be solved. | ||
Though it may never be solved, our [[machine overlords | Though it may never be solved, our [[Necrons|machine overlords]] have reached our level.[[https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol|Athena DeepMind derived AlphaGo program defeated a Go Grandmaster 4-1]]. | ||
===About the game=== | ===About the game=== |
Revision as of 09:02, 1 March 2021
Go is the best game in the (known) world. It was created in the dawn of time by somebody in China under the name Wei Qi. It was exported to Japan in the seventh century as Go, and then to western devils in the 19th and 20th century under the same name. It is also known in Korea as Baduk.
Go is far behind Chess in AI. While the highest level chess players can be defeated by AI, AI for Go struggles even with major handicaps. This is because when the game is played on a standard full sized board, there are far more possible games than in chess, and so even a supercomputer will struggle to find a good strategy through brute force and we have yet to come up with a more efficient AI that works well. So far, only less popular variants with a much smaller board have been solved, and it's possible that larger versions of the game will never be solved.
Though it may never be solved, our machine overlords have reached our level.[DeepMind derived AlphaGo program defeated a Go Grandmaster 4-1].
About the game
Go is a strategy board game involving two players, Black and White. The players take turns placing stones of their own color on the intersections of a grid; common sizes for this grid are 19x19, 13x13, and 9x9 intersections.
Stones are captured by surrounding them. A stone is surrounded when it is not possible to trace a line along the playing grid from the intersection on which the stone sits to an empty intersection, optionally passing through intersections containing stones of the same color. If, after removing captured stones, the stone just played is itself surrounded, the move is suicide and illegal in most rulesets.
To prevent certain deadlock situations, it is illegal to move such that your opponent's move is undone; otherwise the players could reach a situation in which they endlessly repeat the same two positions. In some rulesets it is illegal to recreate any previous position. A situation in which this rule applies is called ko. A typical game can have multiple of these ko-threats. To circumvent a ko-threat, the player must place a stone somewhere else and after that he may continue the ko-situation on his next turn; or ignore it completely and let his opponent take the ko.
If a player believes they cannot improve their position, they pass instead of placing a stone. Once both players pass, they score the game by counting the intersections surrounded by their stones, and then either adding the number of their stones still on the board, or subtracting the number of their stones captured by their opponent, depending on the ruleset.
To offset Black's advantage of moving first, White receives compensation points, typically 6.5 or 7.5 depending on ruleset. The half point is to prevent draws.
If the two players are of unmatched skill levels, White can grant Black the use of handicap stones; in effect, Black places 2 or more stones on the board as her first move, then White makes his (it's traditional to refer to the two players using differing pronouns) first move. In this case the compensation points are only 0.5.
As you might have noticed, Go is a game of absolute strategy where luck has no sway on the course of the game. A player of a year's worth of experience will find playing against someone who has played a month after understanding the basics, even with handicap stones on their side, is akin to going to the playground with a lead pipe and beating up the youngest child with it. There is absolutely no thrill in it (except for some), and you feel bad for poor Timothy.
Also, the entire professional history of the game can be accurately summarized as slowly realizing just how good it is to go first and giving White more and more pity points to compensate; inevitably no amount has proven quite enough.
External links
Board Games | |
---|---|
Classics: | Backgammon - Chess - Go - Tafl - Tic-Tac-Toe |
Ameritrash: | Arkham Horror - Axis & Allies - Battleship - Betrayal at House on the Hill - Car Wars Clue/Cluedo - Cosmic Encounter - Descent: Journeys in the Dark - Dungeon! Firefly: The Game - HeroQuest - Monopoly - Mousetrap - Snakes and Ladders - Risk Talisman - Trivial Pursuit |
Eurogames: | Agricola - Carcassonne - The Duke - Settlers of Catan - Small World - Stratego - Ticket to Ride |
Pure Evil: | Diplomacy - Dune (aka Rex: Final Days of an Empire) - Monopoly - The Duke |
Others: | Icehouse - Shadow Hunters - Twilight Imperium - Wingspan |