OK . . , no excuses . . the VXX long signals were bad. In an effort to determine why, I went back and looked at each of the pair trade signals. Since the PDQ produced a mixed bag of signals, clearly there were some attributes of the correct signal pairs that need to be integrated into the incorrect signal pairs.
My investigation led to a continuation of my earlier post on the zero line rejection behavior of the z-score. While the Z-score rejection signal does offer a useful stop, the salient discovery that will first be integrated into the PDQ is a fixed time stop based on the optimized N day value for each pair.
I spent a few hours manually toggling through each of the PDQ Dashboard performance studies and virtually every one could be enhanced (some significantly) by applying this N day stop. Now these results are a bit surprising (to me) because it suggests that there really is a unique underlying timing cycle for each pair.
While this sounds deterministic, we need to factor in the all-important caveat that the PDQ model is designed to be adaptive and the N day cycle is really only valid as long as the equity curve of the pair trades remains positive.