Designing and Building a Profitable Automated Trading System

Discussion in 'Automated Trading' started by ScottD, Dec 9, 2008.

  1. Here is today's results. I reworked my suppression logic based on ETL's suggestion of using ROC. The results were ok, 3.25 points from 9 trades. The big move in the morning was not caught because the model does not make any trades until after 9:45 (market sync). The work continues! :p
     
    #191     Feb 6, 2009
  2. ScottD

    ScottD

    ETL, I appreciate your post and encourage any further contributions you may wish to make.

    Your charts indicate very nice progress on your ATS. Consider posting some near real time ATS decisions next week. I think that would be a fun thing for people to see.

    In the end, all that matters of course is the profit performance. On the way to profit, the build phase of the Cashcow project strives to be true to Jack's logic schematics. In the background I'm following Jack's lead down a certain design-build project path. At the moment the ATS uses multiple elements of the MACD as rate of change indication and at some stage in the future I could see us incorporating the first derivative of price and other ROC methods, as in your example.

    Sequencing and latching will become more prominent in future releases. The prior release used a certain sequence to define a MACD entwine condition. I didn't have the sequence built precisely and we removed it in this release. As a result we may be over-suppressing in certain conditions.

    Jack and I have some good stuff teed up for future releases and I'll continue to provide progress updates here.
     
    #192     Feb 6, 2009
  3. ScottD

    ScottD

    Nice work WGT. You could consider relaxing this a bit and letting sentiment come into effect as early as bar 2.

    Better yet, bring in the Dow futures/cash price feeds and build a Pre-Sync suppression module. As soon and the pre-sync condition clears, take the next entry signal or take an opening range breakout.
     
    #193     Feb 6, 2009
  4. Thanks Scott. Those are all good ideas and I intend to try some of them. My biggest problem is that since my ATS trading platform is "custom" made, while I have the flexibility to do just about anything I want, it can also take longer.
     
    #194     Feb 6, 2009
  5. Are you making any improvement at all with the iterative refinement ? How does a forward test of the 1st release from 12-10-08 to date compare to the current release from 12-10-08 to date ? Bet it doesn't look any better ...... And your not responding or posting comparisons confirms that hunch !!!!
     
    #195     Feb 6, 2009
  6. ScottD

    ScottD

    I'll make a couple administrative posts to wrap up the first week of forward testing Cashcow Release 1.2 Logic.

    In Feb 5 forward testing, the Cashcow did not take any trades. The entry criteria are a little more limited than the reversal criteria and, on this day, all entry signals were suppressed.

    Corresponding chart attached.
     
    #196     Feb 7, 2009
  7. ScottD

    ScottD

    In Feb 6 forward testing, Cashcow Release 1.2 Logic generated the following results:

    Gross Profit: +2.00 points
    Profit/Range ratio: +0.1
    Trades: 6
    Wins: 3

    --------------------

    Week 1 Summary:

    Gross Loss: -1.75 points
    Trades: 18
    Wins: 9

    --------------------

    Next steps:

    - Validate the two components of new functionality in Release 1.2 -- opening entry logic and automated NSDS lookups.

    - Define and integrate additional logic modules for actions and suppression.
     
    #197     Feb 7, 2009
  8. Padawan

    Padawan

    Mr. Hershey and ScottD, thank you very for your work. The attached document summarizes my current understanding of the beginner's logic. Comments welcome.
     
    #198     Feb 8, 2009
  9. Today the model did reasonably well. 11 trades and 13.25 points. I've stared to make some structural changes to the model to more easily accomodate sequential based actions based on volume and wmcn. The logsheet now indicates the type of volume for each bar. Today's results based on the same fast and slow stochastics and the ROC for suppression used last Friday. I have however started to suppress trading between 12:00 and 1:00.
     
    #199     Feb 9, 2009
  10. Here is model run with the same logic parameters as the ES but on the NQ. The results were decent. 9 trades, 12 points, 0.62 capture ratio, 72% in market.
     
    #200     Feb 9, 2009