Types Of Artificial Intelligence And Algorithmic Strategies In Chess Games

Strategy One: Minimax Algorithm

In this algorithmic strategy, created by John von Neumann. He classified chess as a two-person, zero-sum game with complete information. This means that this class of problems can’t be fully solved using the Minimax Algorithm as the Minimax Algorithm doesn’t go to the depths needed to solve the thousands of different chesses moves. As it has been quoted “There is said to be more moves in chess then there are grains of sand on all the beaches of the world combined” also stated “there is more possible moves in chess then number of atoms in the visible universe”. Meaning there are far too many complex movements and patterns that can happen on a chess board to easily work through with this algorithm.

In a two player zero sum game the Minimax solution is the same as that of the Nash Equilibrium. For games that are two-person zero sum games, there is a value assigned V for one player and -V for the other. This means that if one player has a score of 2 the other must have the score of -2. The name minimax is given because each player is trying to minimise the maximum pay off for the other player while minimising their own loss score.

Strategy Two: Alpha-beta pruning

Alpha-beta pruning is not actually a new algorithm. Alpha-beta pruning is an optimised method to the minimax algorithm that allows us to disregard some branches in the search tree based on their values. This allows us to use the minimax search tree to a better depth, and with faster results. Alpha-beta pruning allows us to stop searching a certain branch if the score is less than the node before. This does not change the outcome of the minimax algorithm, but it makes the calculations faster as stated before. Alpha-beta pruning is faster and more effective if branches with good moves are looked at first.

Defining what alpha and beta are:

• Alpha is the highest the maximiser can be either at the level it’s on you’re a higher level.
• Beta is the highest value that the minimizer can be either on its level or the level above.

Strategy Three: Giraffe

The Giraffe approach is a deep learning technique that is used to play chess. The Giraffe approach uses a self-play methodology. This means there is minimal knowledge given to the program by the programmer. It also uses machine learning but unlike other approaches. This approach doesn’t just use machine learning to perform parameter-tuning on function evaluations that were made by professionals.

Giraffes approach takes it a step further. It does this by letting the machine learning perform automatic feature extraction and pattern recognition. The trained evaluation function performs the same to most of state-of-the-art chess engines that has been made. These state-of-the-art chess engines contain thousands of lines and rows of carefully webbed together pattern recognizers. Tuned by both computer and human chess geniuses, professionals and masters alike. From my understanding the Giraffe approach is the best to this day.