Implementing the Chess Engine using NNUE with Nega-Max Algorithm

The game of chess has been long used for benchmarking algorithms. It is because of the unique combination of intuition, intelligence and memory required to win a game of chess. Therefore, chess has been used to judge the "smartness" of machines. Stockfish, the strongest chess engine available currently, utilises the Min-Max algorithm with Efficiently Updateable Neural Network (NNUE). NNUE being added to the Stockfish repo in 2020, since then NNUE has largely increased the capacity and accuracy of the engine. However, Min-Max algorithm has a peculiar disadvantage with respect to Nega-Max algorithm. That is while Min-Max algorithm computes the feasibility of a particular move only in positive numbers, thus leading to a requirement of large number of comparisons. The Nega-Max algorithm utilises both negative and positive numbers. Thus, Nega-Max simplifies the comparison of feasibility of any two nodes in question. This reduces the overall computational overhead and makes the chess engine faster and more accurate at least in theory. To prove the theoretical efficiency of Nega-Max, we compared "Stockdory", a chess engine implementing NNUE along with Nega-Max, to "Stockfish". Results show that Stockfish outperformed Stockdory while analysing both the unorthodox psychological playing style of Lasker and the more positional play of Alekhine.