You could look at multiple points and try to generate a distribution of when this kind of activity happened. You might find that the number of points results in a pitifully poor confidence level in an expected probability though. Not enough rats and not enough kitchens.