If your data told you that an index would only close down more than 5% ten times out of a period of 1000 trading days, then would you buy it intraday if it were down 7% ?
I ask becaus of your earlier data about closes having predictable characteristics of previous closes, and then the post I am responding to subsequently mentioning bugs.
It could be the case that the bug involves ignoring the possibility that the ten times it closed down more than 5%, it was previously down 7% intraday. That might be an obvious thing to check, but do you have a process for trying to spot such bugs?
Market Data 
Markets 
AI
