Statistical relevence on weekly charts?

Discussion in 'Technical Analysis' started by 99atlantic, May 31, 2006.

  1. How would you go about assessing statistical relevence of strategies on weekly charts? Assuming you capture the actions you want, you're probably looking around 9-18ish trades in a 10-15yr period (atleast I am, heh). Of course it's really hard to assess relevence with such few trades (even if the same type of idea carries over across many different market sectors profitably with only 1 optimizable paramter) but what type of weighting would you give it to make up for the fact that you are looking at fewer bars?
  2. Not sure what you are doing, but I have an end of week system (I don't look at charts, but I use technical analysis - quant type stuff). One EOW system trades on average 4 times a year. The other trades on average 12 times a year. So that's a lot more than your systems.

    But yes, if you have only have an average of one trade per year, you'd need at least 30 years of historical data before you can even start looking at statistics. And then you'd still have to be terribly concerned about curve fitting.

    I have 50 years or so of backtesting and have 223 trades on one system and 670 on the other. But, in reality, I don't look at it on an individual trade basis. I look on it is as a weekly basis. Each week is like a new trade. So I really have about three thousand weeks/trades to analyze.
  3. I'm holding them more long-term - hence why I have fewer trades. Maybe I just need to look at new commodities, haha ^_^