Alternate history?

I do.

I think it is within the context of a larger event that actually happened. Speculation of events is crucial to many areas of history, where, for example, we can fill in some of Herodotus's various mishaps in his account of the battle of Salamis, but to go too far off the mark doesn't really make it academically viable. When one goes so far off the event which they are forging an alternate history for that it's unrecognizable, then it just gets stupid. I would use it as a tool rather than a stand-alone practice. 

In order for alternate history to be a scholarly field one must not bring the alternate history too far. I'd say no more that one hundred years of guessing, because any more and their are just too many variables to get any recognizable history.

What's important it seems to me is to retain the cause-and-effect chain, only restricting oneself to the probable direct results of the original postulated change. That requires considerable analysis of the actual, recorded, history of the period.

Originally posted by TheARRGH A good point. Alternate-history can be immensely interesting and useful (in some circumstances), but you're very correct that to have any real value, it has to have as much precision, reliable information, and thought put into it as possible.

I am not completely confident over what alternate-history is. I assume it is some thing like "what if Hannibal had conquered Rome?". I've seen several book going along those lines (from what if the Romans had discovered America to what if Hitler had won the war) and I can't say I am too interested by it. Still I've been taught in college (and I've used myself) something called counterfactual history.
The principle is the following: two comparable regions were following a similar trend that can be measured. But one of these regions changed something in its system and after that followed a trend different from the one the other region followed. What would have been the trend followed by the region that changed had it not changed? This is basically used to estimate the impact of an event on a region society or economy.
For instance, the rents in the Bronx and in Brooklyn were evolving at a similar pace until the Brooklyn bridge was constructed, as a result Brooklyn's rents hiked twice faster than the Bronx's. What has been the impact of the Brooklyn bridge on Brooklyn's rents? Lets consider that the bridge was never constructed, the rents would have go on at the same pace as in the Bronx and thus would be X% below what they actually turned up to be. Hence the impact of the Brooklyn bridge on Brooklyn's rents equals the real rents with the bridge minus the potential rents without the bridge.
This method is interesting but very complicated because it means that you can account for all the consequences that the modifying events which is difficult at a micro-level and impossible at a macro-level.