Your LLM judge flips its verdict on about a third of pairs when you swap the answer order, leaning toward the answer it read first.

You now judge every pair in both orders and count a win only when both verdicts agree.