new (current) symbols vs backtested symbols

Discussion in 'Strategy Building' started by DTorSwing, Jul 10, 2015.

  1. let me give an example
    say, i am backtesting RSI2 pullback method (a popular one) and it gave a good + PnL results in backtest

    just for example,
    during backtest, it gives symbols, MSFT,SBUK,ARNA,ACAD,ARIA
    during scan today, it gives symbols, ARUN,ABTL,BIOC,BLOX,CLNE

    it showed ++results for backtested symbols
    how can I trust the ++results RSI2 method on the backtested symbols now to trade these new symbols ARUN,ABTL,BIOC,BLOX,CLNE

    or we just test the method (RSI2) during backtest and not for those symbols ?
     
  2. can anyone chime in and comment as I moved from Futures and new to stocks backtesting
     
  3. aqtrader

    aqtrader

    How much co-relation between the tested syms and those new syms? I would say there is no much value to use the tested results for those new syms.
     
  4. oh ok. I have never checked if symbols are co-related, so are you saying they should be co-relation? between backtested and new symbols?

    and how do do i check if they are co-related, like same sector/ sector Industry ?
     
  5. aqtrader

    aqtrader

    Well. not sure how much effective if the corelation is factor'ed in your analysis. I believe it is a good factor anyway to consider. It does not just make sense to me to use un-related syms to try to measure the behaviors of other syms. I also believe it is not enough to just use syms in the same sector/industry, etc to assume they are co-related. I think a quantitative method to measure co-relation between syms is better.
     
  6. fan27

    fan27

    How far back are you backtesting?
    How many stocks are you running your backtest against?

    fan27
     
  7. I was backtesting a set of 3 years and 1 year.. i test a set of 25 symbols at once but I have 20 sets like that in diff sectors
     
  8. fan27

    fan27

    What time frame (daily, hourly, 5 minute, etc.)?
     
  9. 15min chart
     
  10. fan27

    fan27

    Why not expand your backtest with 10 or more years of data? I am afraid you will find it likely does not perform very well. In my research, I found it difficult to find an edge on smaller time frames below daily. Not saying there are no edges on smaller time frames, just that I came up empty handed.

    good luck!
    fan27
     
    #10     Jul 14, 2015