I fear that a noble price doesn't help solving the task of making money in the markets (LTCM) The smarter you are - the longer it takes to make money. Good trading
A problem with cointegration is that the error term is assumed to be N(0,1) ... if I am not mistaken. If this is true, then cointegration pairing is worthless. Say that on T=100, that the residual is 2 standard deviation away. For T=101, there is NO mean reversion. Again, I might be wrong .. dead wrong.