Comparison of Deep Learning Classifiers for the Sudden-Death Positions in Connect6

Board games can be used as a tool for training logic, which has a positive effect on the science popularization education. Connect6 is a very popular puzzle game because it is a fair and highly complex game with simple rules. The objective of this study is to investigate the essential technical challenges involved in constructing a game-based logical learning system for the Connect6 program using deep neural networks, and this study focuses on the comparison of classic deep learning classifiers for the sudden-death positions in Connect6. Our contributions include the followings: 1) Proposing Connect6Net, a deep learning model designed to determine sudden-death positions in Connect6. 2) Experimental results demonstrate superior performance of the Connect6 model with Dense Block technology, compared to its counterpart using Res Block technology. 3) In scenarios with fewer parameters, the model employing SqNxt block technology (C6N-D-s) outperforms its three counterparts.