An Inverse Stopping Rule for Play-the-Winner Sampling