[princecharles]

If I am able to identify an historic trend using a large sample size (to negate volatility), and the trend fails to be explained, even after exhaustive conventional and 'outside the box' theory applications, does this render the voracity and/or value of that trend moot?

I have read it many times here that simple 'bill the cop' data mining (sorry BTC), is at best useless, and at worst a predictive disaster.

Would any of you very bright 'math' guys ever accept that trend into a model with the caveat that sometimes events occur often in this world for reasons beyond our ability to understand?
Or is the rule unbreakable that demands a logical reason and 'in the lab duplicitity' for an historic trend to be considered seriously as a useful predictive component?

Really interested in the tank's take on this.

[MonkeyF0cker]

Divine intervention? What could possibly be inexplicable?

[mathdotcom]

The answer is you obviously don't know enough about regression analysis to be using it. I'd otherwise try to give a more useful reply but that's the culprit here.