Cuckoo hashing for repetition detection... no gain?

[chessbit]

Hi,

I have just implemented the cuckoo algorithm to detect a repeated position before making the move. On paper, it sounds like a clear win. But after several thousand games of selfplay, I come to a perfect draw against the version without it. So maybe I just have a bug that I did not see (though I know the detection is triggered).
Did anyone here have success with it? How much elo do you estimate to have gained?

Regards,
Thomas