# Last edited on 2014-05-29 13:42:54 by stolfilocal
# NOT SENT

[quote author=aminorex link=topic=178336.msg6876332#msg6876332 date=1400773528]
This is not blind extrapolation, but frequentist statistics applied to a structural model.  [ ... ] In general it is not wise to argue "I am ignorant, therefore everything is unknowable".
[/quote]

Well, you must agree that in order to predict the future behavior of something, it helps to find out the causes of that behavior.  Then, if the ultimate causes are stable, one can try to model their effect statistically and use past frequencies as probabilities for the future.

For the weather, the ultimate causes are geophysical phenomena (rotation and orbit of the Earth, water cycle, etc.) and [i]for their nature[/i] can be expected to remain unchanged for millions of years.  With that basic information, one can build a stochastic model incorporating 1-year and 1-day periodic factors (but not a 17-day factor), plus random factors; and use frequentistic analysis for the latter.  More sophisticated statistical analysis of the data can give only limited improvement in prediction accuracy; to get beyond that, one must try to get a deeper understanding of the phenomena and/or more data -- such as wind speed and direction, and the current weather a few hundred miles away, all around. 

Back to bitcoin, the first step should be to understand what causes the significant price changes and trends.  The bad news is that the main causes are currently in China, and (apart for the Feb--Apr gradual downtrend, and stable prices during holidays) seem to be unpredictable even to the Chinese exchange owners.   That knowledge tells us that (apart for those two things) trend analysis willl not work asl long as China defines the price.   It also tells us that, in order to predict what will happen if and when China becomes irrelevant, one must try to remove the Oct-Nov rally, and maybe even the early 2013 one, from the price history.  Then, if one believes that the remaining causes are persisitent and regular, and there are no really new causes, one may use extrapolation on what is left.