Tail estimates motivated by extreme-value theory