Programable vs. Non-programable edges?

Discussion in 'Strategy Development' started by Zr1Trader, Feb 5, 2011.

  1. The only type of statistical edge I can find is razor thin after commission/slippage and probably not scalable. It's automated and took me a year to figure out how to make.

    How in the hell do you know if you have a discretionary edge? I've done so many hours of back testing automated strategies only have a couple things that work. Manuel back testing on the other hand takes so long only to find out if each idea I try works ( they never do )

    Countless hours of studying charts /back testing manual ideas and it seems that nothing is working. I must be retarded because I've been at this for over 3 years .

    I've heard that the best edges are not programmable , is this true?

    What should I focus on Discretionary or System trading?
  2. Cheese


    Amateur backtesting, a favourite at ET, produces little or nothing in usable results. An input of OHLC is too primitive but let me suggest further that 'divide and conquer' filtering has not been conducted first.

    Statistical sorting derived, under certain parameters or filters, from the daily trading figures (of at least 2 years data), provide remarkable and highly valuable correlations/indications for each trading day ahead.

    The point of this post is to be helpful. A very usable answer can be found but not from the endless backtesting conducted by various ET members.