Economists and psychologists have documented patterns of individual decision-making behavior (e.g., loss aversion) whereby losses and gains are treated differently. However, there has been little evidence of such patterns in multiplayer games. I report results showing the strongest evidence I know of that this phenomenon is present in games. Experimental subjects play two hawk-dove games that are identical up to a constant; in one, all payoffs are positive, while in the other, payoffs are negative if and only if both players choose hawk. Under both fixed pairs and random matching, differences between the games are substantial, significant, and consistent with loss aversion.