OK, I understand. It was not really clear from your first post that you were doing a Wald-Wolfowitz test. Then I do agree with your results. It is the fact to find 1332 "runs" that is 3 sigmas and not the difference between up and down events. My fault, I misread you.
However, the problem...