Hi, why didn't you model the spread with a cointegration coefficient? Wouldn't this be more theoretically aligned? And how important is the parameter constraint of C < D when dealing with daily price data really? I noticed that when I relax this constraint, many more pairs pass the constraints (0 < B < 1, C >= 0 and D > 0).
Hi, why didn't you model the spread with a cointegration coefficient? Wouldn't this be more theoretically aligned? And how important is the parameter constraint of C < D when dealing with daily price data really? I noticed that when I relax this constraint, many more pairs pass the constraints (0 < B < 1, C >= 0 and D > 0).
I believe I modelled the spread as it was described in the paper. There is a link to it in the beginning of the article.
If you have C>D, then the underlying process is too noisy, which can make it difficult to generate reliable trading signals.