I'd believe it. I had a similar bug in a minimax AI I made to play Hive. I had messed up the game-end valuation in such a way that the AI basically valued winning less than anything but outright losing. It would play a perfectly normal and competent game right up until the end, at which point it would try its hardest to draw out the game indefinitely, usually by forcing me into terrible positions that nonetheless didn't particularly benefit it at all.
(Like blahedo points out, this is possible because, unlike say basic MCTS, minimax uses a heuristic to judge board positions prior to game end, so it is possible for the two metrics to be out-of-sync.)
(Like blahedo points out, this is possible because, unlike say basic MCTS, minimax uses a heuristic to judge board positions prior to game end, so it is possible for the two metrics to be out-of-sync.)