Emulation of humanlike behaviour in chess: An optimization of an MTD-f search based chess engine

The main objective of the research was to code a chess machine using the Python programming language, which would make as human-like moves as possible in chess situations. We approached the problem in a novel way: instead of using neural networks, we would utilise classical engine architectures to develop the AI. We built three different machines by modifying the search algorithm of Sunfish, a chess engine. Adjustments were made based on the performance of the engine in positions that humans had played in. We found that the search depth had a significant positive correlation with the Elo level. In the end, our developed AI did not play in a more human manner than the original. We conclude that our method of optimisation is not the most effective compared to alternatives.