The Evaluation Function Explained Further

In 1950, Claude Shannon introduced two basic strategies for limiting the number of possible moves to be considered in chess. Both techniques, with minor adjustments, are still utilized presently by both credible and bogus computer chess programs. Here we will look at the two techniques in-depth.

The Type A Strategy

The first technique considers all possible moves up to a limited number of plies. At this stage, criteria are utilized in order to select the most favorable move based on the board positions obtained at the conclusion of the search.

The contention is that, since it is unlikely that all the possible moves and counter-moves can be used until the duration of the game, the computer limits the number of plies to a certain number that makes it possible to respond in a certain period of time.

It is interesting to note that human players likewise consider a certain number of plies until a favorable position is obtained. Of course, the computer would need some technique in evaluating the positions obtained after the sequence of plies is determined and choose the most favorable sequence among them. The evaluation function designates for each position a number which can provide the best advantage to a player. Whether or not, the player can take advantage of the position will depend on the evaluation function.

The Type B Strategy The second technique developed by Shannon reduces the number of moves to be considered per level. With the help of a move generator, a set of favorable moves for each position will be determined. Furthermore, the number of levels of the tree is not fixed. Generally, the minimum level of depth is evaluated even further. If the position is unstable, it is further examined until a desirable level is reached.

It is easy to understand such reason. Let's say that the queen of a player captures the rook of the other player. When the current position is evaluated by the computer, it is chosen because of its material advantage, even if, in the ensuing move, the queen may be captured by a pawn which could be a disadvantage for the player.

It is interesting to note that an extensive search is not effective if only a few moves will be chosen for every position. As a result, computer chess programs developed in high speed machines have the tendency to be class A strategist while those that run on microprocessor architectures tend to be Type B strategists.

Archive


Meta