Antirandomizing the Wrong Game

We study a variant of the tenure game introduced by J. Spencer (Theor. Comput. Sci. 131 (1994), 415-429). In this version, chips are not removed from the game, but moved down to the lowest level instead. Though the rules of both versions differ only slightly, it seems impossible to convert an upper bound strategy into a lower bound one usingthe antirandomization approach of Spencer (which was very effective for the original game and several others).For the upper bound we give a potential function argument (both randomized and derandomized). We manage to prove a nearly matching lower bound usinga strategy that can be interpreted as an antirandomization of Spencer's original game.