ABSTRACT
In the 1990s, as computers got faster and cheaper, the search for Hex strategies shifted from computer-assisted human efforts to human-assisted computer efforts. In 1992 in the Netherlands, Jack van Rijswijck was looking for a project for his master’s thesis. A professor suggested that he try to get a computer to play Go. Van Rijswijck was hooked: for his project, he and a classmate used reinforcement learning to try to teach a neural network to play Hex. From the beginning, as Hex players realized that first-player winning strategies must exist, they sought to find them. As Piet Hein told Politiken readers in 1942 and John F. Nash explained to Martin Gardner in 1957, finding such strategies on small boards is easy. In 2001, Ryan Hayward started a project to build the eponymous Solver for Hex. Yngvi Bjornsson wrote the initial version, helped by Mike Johanson and Nathan Po and Jack van Rijswijck.
