This paper examines games with non-neutral option labels (such as "A", "B", "A", "A") and nds surprisingly invariant behaviour across games. The behaviour closely resembles the choices people make when they have to bet on one of the options in individual lotteries. An option's 'representativeness' (lack of distinguishing features) and 'reachability' (physical centrality, salience, and valence) determine choice behaviour in both the lotteries and the highly strategic games. There is no evidence of people best-responding to others' betting(-like) behaviour. This is in line with the idea that once people decide that strategic reasoning would not take them any further, they pick an alternative as if they were betting on one of their 'current best-responses'. The ndings explain the well-documented seeker advantage in hide-and-seek games, as well as why participants often display behaviour that could be exploited by others. On top, they help understand why in national lotteries, people also tend to bet on identical subsets of the available numbers.