How Computers Determine The Best Possible Move In Chess

The existing state-of-the-art technology in computer chess is quite complicated, but all of them consist of blind calculations that are very basic at the root.

Supposed you have a chess board prepared for a new game. Each player has 16 chess pieces. White starts the game and has twenty possible combinations of moves:

A move can advance one or a couple of squares. Either knight can be advanced in a couple of varied ways.

The white player opts for any of the twenty moves and executes it. The player with the black pieces has similar options so they can make a move.

The succeeding options of white is determined by the prior moves that they made but similarly there are twenty moves available to them in the present board location. The same thing holds true for the black pieces.

This is how computer chess works. It looks at all the possible moves and generates a tree for all these movements.

The entire movement tree is not calculated by the computer. It only attempts to generate ten or twenty possible positions. Let's say that there are 20 probable positions. In a 5 level tree, there are 3,200,000 positions involved. In a ten level tree, there are 10 trillion possible positions. The extent of the tree that can be calculated by the computer is determined by the speed of the computer involved in the game. The fastest computers can build and analyze up to millions of positions in a second.

As soon as the tree is generated, the computer would then analyze the board positions. This means that the computer would check if the layout of the pieces is "favorable" or "unfavorable." This process is known as evaluation function. The basic function might just look at the number of pieces available to each side.

Let's say that the computer is the white pieces and the current setup of the board has 11 white and nine black pieces, the basic evaluation function is:

11 - 9 = 2

For computer chess, the formula is basic because there are some pieces that have higher value than others. Thus, the formula might add some value to each piece.

Evaluation function makes things more complicated by including issues such as board position, domination of the middle board, susceptibility of the king to check and of the opponent's queen as well as other parameters.

Regardless of its complexities, evaluation function is reduced to a single digit that represents a "favorable" board position.