Different time interval performance comparison

Discussion in 'Technical Analysis' started by marketnoize, Sep 21, 2002.

  1. If I take the same system, and first backtest against a one minute bar data set, and then backtest it against an end of day data (both data sets of equal size and starting point), can I make a direct comparison of the two performance numbers?

    Or will I have to expand the one minute bar data to be of the same timespan as the end-of-day data set? If I have 10 days worth of historical, end of day data, in order to make a just comparison do I need just 10, one minute bars, or do I need 3,900 one minute bars (10 days * 390 one minute bars in a day = 3900)

    With all of my tests thus far, I have only utilized end-of-day data. But I am looking to get more realistic numbers. Since my automated trading system can enter/exit throughout the day, these numbers are approximations at best.

    Thanks for any insight.
     
  2. No, the results will differ.

    It's the same as putting a daily chart signal to a weekly chart signal(maybe monthly would be better). Very different.

    It also depends on the kind of signal you use. Volatility based will differ. Moving Averages will differ also. Intraday moves more rapidly compare to daily charts. You'll have a consolidation and suddenly moves. It's a lot like stairs. As I recall, I thought breakouts are pretty much OK, still you'll have some major slippage compared to daily.

    Just my observation.
     
  3. chase_jo

    chase_jo

    No, you don't understand the question.

    I understand that the results will be different; with such different data sets one would have to anticipate very different results.

    But if a system has an ROI of 20% with daily data and an ROI of 70% with 60min bars, am I overgeneralizing the problem by stating the system performs better with the 60min bars? Or should there be some sort of data scaling before making the comparisons?