I think the movement works correctly if you accept that tigers can't go back to their previous spot, which is rough. However, the bug seems to be that sometimes

I think the movement works correctly if you accept that tigers can't go back to their previous spot, which is rough. However, the bug seems to be that sometimes, when a tiger TRIES to do the above, it will trigger a loss for the tigers even if there are legal moves. I have managed to get a win against Hard Goat AI. Generally you want to be positioning tigers with 2 spaces between them, and when you need to break the pattern, do so with a tiger that can get back to its original space in a right triangle pattern with the original spot being the right angle vertex. I think this is mostly just possible because the AI plays imperfectly. Again, if you forget which tiger was last positioned where, you'll probably trigger a loss for yourself.