There is a very limited set of circumstances in which playing a game with negative EV can be rational (e.g. if your utility function for money isn't linear); everything else seems like a more or less thinly disguised attempt to exploit certain properties of the human mind's/brain's reward system. (It's a philosophical question on whether consent can even theoretically be given to that exploitation.)

I'd argue that the magnitude of the expected loss per game (i.e. everything or only a fraction of it – but in iterated games it's always everything as well) is much less relevant than that.