Nature of the predictions genre


The key variables that differentiated the titles that had been number-one bestsellers from the titles that had failed to be were whether the title was literal or figurative, the word type of the first word, although it was interesting that the word type of the second word didn't matter, and also the title's grammar pattern.
it was nearly 70 percent perfect, which is 40 percent better than random guesswork, which, given the fact that it covers such a long period and that tastes obviously change and various other methodological difficulties, was actually a more successful model than we expected it to be. One of the surprise findings was that the length of the title did not seem to matter. Cherished publishing industry folk wisdom suggests that a best-seller should have a short title, if not just one word, but we found that the number of words in the title didn't affect its sales prospects. And another interesting finding was that titles that are figurative or metaphorical outsell those that are literal.

This sounds interesting but I wonder how interesting it really is. All sorts of gerrymandered theories for the last sentence (my emphasis) but in fact there is no basis there for a coherent theory. Instead it is always interesting to be reminded of two things:

1. There are two perspectives on predictions. One is their ability to actually predict a single instance of some future behavior (which is very difficult) and one is to express trends in past behavior in terms of probability. Much of the frustration with inaccurate prediction stems from confusing the two.
2. Statemetns about predictions like these are in fact a ritual act that connects much of what we know into wholes both through confirmation and disconfirmation.

From a scholarly perspective, the actual content facts behind the study are of no interest whatsoever.