Yes but that's still using the past. All we have to work with, in order to formulate scientific theories, are correlations and patterns detected at some point i

Yes but that's still using the past. All we have to work with, in order to formulate scientific theories, are correlations and patterns detected at some point in the past. Of course, some correlations and patterns are much stronger than others, and we can be more confident in the predictions they yield. So really the fallacy is about using too weak a model to predict the future.