Under what scenarios would you consider some indicator data untrustworthy?

Discussion in 'Data Sets and Feeds' started by Rocko Bonaparte, Feb 1, 2010.

  1. This could be pretty broad so for now I will focus on a problem I'm having. I have been trying to develop an MACD-based trend following system. When I run it on individual charts it looks good. It blows up when I run it across 8,000+ equities though. I look at the picks it makes and see a recurring problem. I had this happen when trying to do a mean reversion system too using MACD.

    Generally, the indicator would just become screwed up if the stock completely crapped itself from one day to the next. Say, the two bars don't overlap at all. Not even close. One would assume some news item came out that completely screwed up the stock. It just started the day 50% lower. You can imagine a mean reversion algorithm--not knowing what is really going on there--going off like crazy. But it also throws off stuff like trend following I had developed.

    I'm trying to buff up the moving average code I have for the base part of the MACD code. Generally I will put in some rules that if, say, two adjacent bars don't touch at all over, say, a 25% range, then the moving average acts like it's starting over for the time after that gap. So a 9-day EMA, seeing a big gap, will reset itself and then have nothing to say for 9 days following while it basically learns the new situation.

    This 25% rule is completely arbitrary so I thought I'd ask on here some thoughts, and considerations for other scenarios anybody wanted to share where their indicators barfed on them like that.
  2. mikasa


    it sucks to be a new guy

    trading according to signals from public indicators has no edge

    even if you took opposite trades you would lose money

    do some readings

    try to waste as little time as possible because it will take you 10 years to learn it

    3 years if you are smart

    2 years if you are super smart

    5-10 years if you are average

    good luck