Just a two cents worth here. :) Chess itself might perhaps not be too mathematical, but the chess evaluation functions of any chess-playing computer program seems like a mathematical object. After all, these are maps from the set of chess positions to $\mathbb{R}$ and they are bound to satisfy various properties. Given any two chess programs that are both strong and might be expected to be decent (in terms of current technology) approximations to objective truth, one might probably expect them to be "close" in some meaningful way that one could perhaps attempt to define.