Computer Chess: The History Of Computer Chess

Computer Chess

History of Computer Chess

The time period of 1949 and 1950 is considered to be the birth of computer chess. In 1949, Claude Shannon, an American mathematician, wrote an article titled “Programming a Computer for Playing Chess” (5). The article contained basic principles of programming a computer for playing chess. It described two possible search strategies for a move which circumvented the need to consider all the variations from a particular position. These strategies will be described later when we talk about implementing chess as a computer program. Since then, no other strategy has been developed which works better and all engines use one of these strategies at their cores.

About a year later, in 1950, an English mathematician Alan Turing (6) (published in 1953) came up with an algorithm aimed at teaching a machine to play chess. Unfortunately, at that time there was no machine powerful enough to implement such an algorithm. Therefore, Turing worked out the algorithm manually and played against one of his colleagues. The algorithm lost, but it was the beginning of computer chess.

In the same year, John von Neumann created a calculating machine which very powerful for the time. The machine was built in order to perform calculations for the Manhattan Project. But before it was used there, it was tested by implementing an algorithm for playing a simplified variant of the game (6x6 board without bishops, no castling, no two-square move of a pawn, and some other restrictions). The machine played three games: it beat itself with white, lost to a strong player, and beat a young girl who had been taught how to play chess a week before (7).

In 1958, a great leap in the area was made by scientists at Carnegie-Mellon Univ...

... middle of paper ...

...age here is that, it is possible to accumulate information about the effectiveness of each move throughout the whole game tree, unlike killer moves where only a certain sub-tree is considered.

Each time a move proved to be good (caused a quick cut-off or achieved a high evaluation score), its characteristic, which indicates how good this move is, is increased and the greater this characteristic is, higher is the move’s privilege in the list. For example, the move that was placed among the best ones 2 plies ago will still have a good characteristic and can be placed at the top even if a different piece can move so now. Thus, in Fig. 4, after the game continued 1. … Qxc3 2. Bxc3, white’s move Bxf6 (instead of Qxf6 a move ago) is still dangerous. Of course, all this makes sense only for a certain period of time, and so the history table must be cleaned periodically.