An antiprediction is a statement of confidence in an event that sounds startling, but actually isn't far from a maxentropy prior. Antiprediction fights the tendency of people to clutch to the last possibility, even where it's completely implausible that a certain event will happen, to a point where thinking about it becomes counterproductive or otherwise distorts the decision-making process. Just as prediction simplifies reasoning by allowing to assume that an event will occur and only planning for the case where it does occur, antiprediction simplifies the reasoning by allowing to completely disregard the case where the event occurs.
The reasons for keeping a hypothesis salient despite its unlikelihood are many; it may be emotional attachment, availability created either by experience or problem framing, or any other cognitive bias.
Often, it is sufficient to see the privileged possibility as but one among many equivalently probable events, and use maximum entropy principle to divide probability equally among them, leaving each very little. For example, each of the millions of possible lottery combinations are equally likely, and so any one person's ticket is very unlikely to come up: winning the lottery should be seen as one possibility among millions, not one of the two possibilities of "winning" and "not winning". Instrumentally, it's better to believe that winning is impossible, than that it's likely, if the actual probability is very low.