(OP's language was C) only one game was run in the random tests, due to the peculiar setup of this kata in C (the function has to be stateful). I rewrote the tests to run 100 games, several solutions have been invalidated.

Maybe add some fixed test for edge cases to avoid depending on luck with the random test cases.

I still managed to pass with buggy code with only a couple of attemps.

the logic for the bounce back in the funktion move is false: Instead of moving the piece back it places it exactly on the last square.

The random test cases did sometimes not covered this case - that's why I was able to post this "solution" :C

(c-version)

there is a typo in the provided test: The output message labels both expected strings as "S2" - should be labeled as "S1" and "S2".