Building and sharing a profitable strategy

Discussion in 'Strategy Development' started by frostengine, Nov 17, 2011.

  1. Over the years, I have been asked to be more open about my strategies. In this thread I will be detailing a new strategy I am building from scratch and completely open to the public. All aspects including the final source will be available.

    I am hoping for a lively discussion about the direction of the strategy (both good and bad) to help improve the overall strategy. The discussion will take place here, whereas the details, images, and source will be available via my blog.

    To get started, read the basis on which this strategy will be built found here <a href="">Strategy</a>

    Essentially, the first step was to identify various entry conditions while still maintaining an edge over random entry. Next, we will look to filter and tweak these entries. Currently, there is not much to it. But over time as more advanced aspects of my strategy design comes into play such as neural networks and some unique data mining techniques, we should see a good strategy evolve.

    Looking forward to a good discussion and hopefully a profitable strategy by the end.
  2. Its been several days since I last posted. This thread is not DOA. Just taking some time to fine tune the next steps. Building a strategy is not fast. Especially when utilizing my custom pattern recognition software which sometimes diverts my attention. Had a bug to work out past few days.

    Essentially the next stage of this process is to have the pattern recognition software look for patterns that lead to success or failure. Then weed out the losing trades
  3. Raising cattle so you can eat them later?
  4. maler


    If your profitable, you take another trader's money.
    If you tell the other traders how you are going to take their money,
    you will find it hard to do so afterwards.
    Either because your mark fixed his game,
    or because a faster trader beat you to the mark.
  5. jb514


    I've heard ninja trader's back testing is not too accurate. But 58% win rate with $8 average trade doesn't sound too good, especially with a $4,000 drawdown. Have you tried forward testing at all?
  6. I've counted over 20 parameters in this strategy, which is way to many, in my opinion. In fact, I believe that any system which has more than 5 parameters is unmanageable. In regards to performance, from my real life trading experience, the spread/slippage/commission overhead on the ES is about $20 per round turn. So if your system makes $8 per trade before accounting for this overhead, it's a certain and consistent loser.
  7. 20 parameters on 13k trades is about 650 trades per parameter. This is a pretty good number.

    A more realistic figure is about $15 per RT on slightly longer time frames. However, average trade was at $6 not $8 on the outlined initial strategy. So yes, it's not profitable, it's making too many trades to counter the spread/slippage and commission.

    I'm not sure how you can openly develop a strategy. In order to get any interesting inputs from somebody else they would have to have full access and knowledge of your methodology. Also, as someone pointed out, NT isn't that great at realistic backtesting.
  8. Multicharts already much better since can test on bid&ask.

    And a PF 1.2 strategy I dont consider developable.
    Sorry, I dont have anything constructive to add.
  9. I think the main issue is this was meant to be a starting point. The strategy is still using 5 bar exits for crying out loud. First identify an edge, then you look for ways to milk that edge until profitable. More thought was put into getting the # of trades high than anything else. Using that as the basis, it can now be filtered down to something profitable.

    You do not build a profitable strategy all at once, I am trying to show the steps you must go through. Once piece at a time.
  10. There is an assumption here is that to build a profitable strategy, one "must" start with a large number of trades and subsequently filter them. First, that's not how I build my systems. Second, in can be demonstrated that starting with a system that generates trades randomly gives better results than your starting system. So, what's the value of that first step?
    #10     Dec 16, 2011