A point well taken. The 13 Keys were developed in 1981. They retrospectively predicted every presidential election from 1860 to 1980, and prospectively predicted the elections from 1984 to 1996, with a near-miss for 2000.
I agree that a lot of the criteria are subjective, but I don't think that's avoidable. This seems like a pretty good measure, based on its track record -- and the analogy to snowfall in Peru is not really on point, since there's no plausible causal link between snowfall in Peru and American presidential elections, but it is plausible to think that the factors listed in the 13 Keys would have an impact on the presidential election.
I think rather than slamming the 13 Keys, it is more interesing to look at why they failed in 2000, and see if it would be a good idea to tweak them. For that, we really need more data, and presumably the 2004 election will provide us with that. If 2004 results in another failure, we should be very skeptical of the formula; if it is a success, then we might conclude that the formula remains a highly accurate, but not a perfect, predictor.