[varsigma (OP)]

One Vladimir Keilis-Borok, a Russian geologist, developed a theory for predicting earthquakes which, when he collaborated with a colleague, Alan Lichtman, became a theory that predicts the outcome of US presidential elections.

So that has to be an example of looking for a pattern somewhere and then seeing the pattern somewhere else entirely. And it just so happens it's easier to predict a presidential race in the US than it is, still, to predict an earthquake, even in California.

Why though, to all the above? Why does it work or succeed, but fail in its original domain, this pattern-recognition theory, also why is Alan Lichtman a household name, but the other guy isn't?

[varsigma (OP)]

What's different about earthquakes and elections? One fairly obvious difference is that elections, at least in the USA, are held on a regular 4-year cycle.

Earthquakes don't do that, so I think it's safe to say the theory of presidential predictions is a simplified version of the earthquake prediction version. With a fixed cycle it's much easier to "distill" the factors with the greatest influence on the outcome and test the theory against historical results.

That last piece, the historical results, are less statistically meaningful also with earthquakes, because of the lack of longterm, consistently gathered data.

[alancalverd]

Thanks to the corrupt system for electing US presidents, the office only alternates between two parties and recent history shows that it is unusual for either party to hold it three times in a row, so if you simply draw a best-fit sine curve with a frequency of about 6 years your next guess will be more than 50 percent correct!


Problem with earthquakes is defining the thing you are trying to predict. Richter 1, 2 or 3 quakes are of no consequence and happen all the time, all over the place. The bigger ones are significant, mostly localised along known faults, and necessarily at random intervals and with variable intensity. So what exactly do you mean by an earthquake?