Tests on the history of cointegrated pairs.

I think that history testing of preselected pairs for a pair trading strategy is the most blatant kurvfitting.. Normally, take pairs? Check the cointegration of two shares and if there is one, create a new synthetic instrument from them — couple. And then they test this pair on history. But this is self-deception. This is the same as if you calculated the ten highest gains over a period and began to test the buy-and-hold system on them over the same period.. Result, for sure, obvious to everyone without any tests…..