Computer chess

Computer chess refers to both specialized hardware and software designed to play the game of chess. These tools serve many purposes for players, including practice, entertainment, and deep game analysis. Modern applications can play at the level of a chess grandmaster or even higher. Such powerful engines are available on everything from massive supercomputers to small smartphones.

To make decisions, computer chess uses heuristic methods to navigate the game. A heuristic is a practical method used to find a good enough solution to a problem. The computer builds and searches through mathematical trees that represent sequences of moves. Each point in this tree is called a node, and these trees can contain millions of nodes.

Modern computers are fast enough to process hundreds of thousands of nodes every second. They use extension and reduction heuristics to narrow the search to only the most relevant moves. This process allows the machine to execute the best possible sequence of moves during a game.

There are several ways to interact with chess software. The most common type is a program that simply plays the game. A human makes a move, and the engine calculates a response. Often, the engine and the Graphical User Interface (GUI) are separate programs. The GUI is the visual part of the software that shows the board and pieces. This allows players to choose different board styles or even 3D animated pieces. Some engines, like Stockfish, rely on neural networks that run on a computer's CPU. Other engines, such as Leela Chess Zero (Lc0), use networks that depend on GPU performance.

Software can also be used for specialized tasks like training or database management. Players use databases to search through historical games and check their own statistics. Programs like Chessbase are common for professionals, while others like Scid work on Windows, Mac, or Linux. There are also many training apps available for mobile devices. Some programs, like Fritz, include a "Friend Mode" to match a player's skill level. This uses parameters to reduce the engine's Elo rating, which is a measure of its playing strength. This makes it possible for humans to practice against a challenging but fair opponent.

The history of computer chess began in the 1950s with software on vacuum-tube digital computers. These early programs were very weak and easily defeated by beginners. In 1957, researchers discovered refutation screening, which uses alpha-beta pruning to optimize how moves are evaluated. Despite this, many experts in the 1970s believed computers would not reach Master level anytime soon. In 1968, International Master David Levy bet that no computer would beat him within ten years. He won that bet in 1978 when he defeated Chess 4.7.

The field changed rapidly in the 1980s as processing speeds increased. The program Belle began defeating Masters by examining 100,000 positions per second. Experts noted that much of this improvement came from raw speed rather than better logic. By the 1990s, the gap between humans and machines began to close significantly. In 1996, IBM's Deep Blue became the first computer to win a game against a reigning world champion, Garry Kasparov, under tournament time controls. In a 1997 rematch, Deep Blue won the match with a score of 3.5 to 2.5.

Today, the landscape of computer chess is very different from the early days of AI. In 2006, Professor Monty Newborn declared that the science of computer chess was completed. The field was once called the "Drosophila of AI" because it was used to study the edges of knowledge engineering. Now, playing chess with a computer is considered a mundane computing activity. However, solving chess—finding the perfect move in every possible position—is still not possible. The game has an extremely large number of variations that even modern computers cannot fully map.