Suggestions to Improve Automated Trading...

Discussion in 'Automated Trading' started by greaterreturn, May 4, 2008.

  1. tommaso

    tommaso

    Hi Wayne,

    we are talking almost of the same thing. By "trailing lock" I mean practically a trailing stop limit, with the notable difference, as you also indicate, that the stop begins *only after* a given (high) profit threshold is reached.

    That's why I call it "trailing lock" because we are not actually "stopping a loss" but, somehow, locking a profit, which has already been achieved in the past time (although not immediately cashed).

    As you rightly note this is a smart way to do it (and probably the only reasonable one), because by postponing the time when you "take profit" you allow the profits to run, if they like to.

    Another important difference respect to a classic trailing stop limit is that, as we control it programmatically, we can allow the tolerance to grow with profit. And this is a fundamental mechanism to be able to ride trends or movements of larger timeframe.

    It's like a car driving, where we start driving a 500cc engine and we end up sitting on a 10000 cc.

    That's also why the more we stay in the market, the more we should get.

    Let me know how those ratios work for you.

    Tommaso

    PS
    No doubt ordinary additional stop orders are needed in any case, as you say, as a safety measure against long disconnections.
     
    #41     May 22, 2008
  2. rhoniel

    rhoniel

    Hi Wayne,
    Its great to see that you are testing exits on random entries. Can I ask you a couple of question?

    1) I have noticed that when I use fixed SL and TP only about 27% of my random entries result in winning trades. On your winning strategy based on random entries, what percentage of you trades are winning trades?

    2) On your winning strategy based on random entries, what kind of SL are you using? I mean whether you are using ATR based stoploss, some other TA based stoploss or just a fixed price based stoploss limit?

    Thanks for your contributions in this thread.
     
    #42     Jun 3, 2008
  3. I'm buildling tick processing engine so I can test strategies on 8 years of tick data and multiple pairs simultaneously.

    I'm just using Excel or stuff for charting.

    But I'm not at the point of doing some testing.

    One thing interesting I learned that high frequency randing trades with simply a SL and TP result in a slight average profit.

    It randomly generates 20 signals during each day (8 am to 12 noon EST) of either long, short or flat.

    I tested a variety of SL and TP using an "optimizer".

    The trading results were about 47% winning with a slight average profit.

    But surprisingly, that makes the DAILY winning % OVER 50% profitable.

    And most fascinating, that makes the weekly winning rate 100% percent. I said ONE HUNDRED % profitable winning rate.

    This convinced me to focus strategically on the winning % per week or month rather than the trade-by-trade %.

    Really, who CARES about the winning % of trades themselves if they provied weekly or monthly income that's steady and reliable?

    I'm now working on improving the profitability with by adding simple entry rules that are less random.

    First I'm going to keep the timing randing but choose the direction based on a trend or something like that.

    Then I'll try different techniques for timing the trades.

    Wayne
     
    #43     Jun 3, 2008
  4. Wayne, are you up for some input that may be disturbing?
     
    #44     Jun 5, 2008
  5. Sure!
     
    #45     Jun 6, 2008
  6. nbeeyard

    nbeeyard

    I'm new to this forum. I'm french and i couldn't find such exhaust forum about trading in france. I'm surprise by professionnalism, experience and passion of people who write.

    I'm interesting in automatic intraday trading system for 1 year.
    I just finish to develop a plateforme with Matlab for almost all operations : retrieving datas from my broker by parsing webpages (every 30s), backtesting, and applying strategy.
    I just use windows macro instead of informatic code which perform HTTPS protocol in order to update webpages, and to sell or buy. Finally i have a system which has zero cost for me, i don't need any abonnement, any API,...


    For the moment i have only 1 simple strategy that i tested to not enough datas, but with this system i'll have more and more "ticks" with time. I just paper trade now and i'm waiting 1, 2 or months until i hope to trade with money (depends on result of this strategy, may be i'll have to take more into account remarks of this forum)

    Contrary to what i read, i only focus on trading kind of options (look like pull or call but with more diversity of behavior and quicker maturity) proposed only in some countrys in europe by one bank. So i want to only analyze options evolution and add some parameters such as local volatility for this analyze.
     
    #46     Jun 7, 2008
  7. I can see three possible problems with this approach:

    1) Overoptimization problem: you optimized SL and TP over some fixed period. Did you test the best found SL and TP combination in out of sample data set? What were the results?

    2) Relying on fixed values. This is invalid approach, because volatility changes. Robust approach would be to use volatility units instead of fixed number of points or pips. For example, you could use 5% of average range of past 10 days as a volatility unit for a given day.

    3) Did you include spread costs into this testing? You say it has a slight average profit and high frequency. Here's a headache for you: suppose your average profit is 4 pips per trade. You include spread of 3 pips and your average drops to 1 pip per trade, that is, 75% of profits gone. If your average is 3 or less pips per trade, then you can effectively put your strategies into a trash can.
     
    #47     Jun 28, 2008
  8. MarkBrown

    MarkBrown

    who cares about the trading method, "IF" you have one good for you.

    i have advice - you better be working on hardware, software, data and all other plethora of things that are going to go wrong!

    one day the reaper he will come and take more than you have. i have spoken.

    mb
     
    #48     Jun 30, 2008
  9. Based on suggestions in this thread and my 20+ years of software experience, I'm building my own auto trading platform and making solid progress.

    At the moment, I have a simple trend strategy that gives me a signal of 1,0,-1 of whether the trend of the instrument is long, flat, or short.

    That serves as input to other strategies I'm testing now.

    I plan to implement a long counter trend, a short counter trend, and a chop strategy driven by the trend signal.

    Of course, I consider slippage and commission because for the trend strategy, the average trade is 37 pips with 50% win/loss and only 85 trades in 5 years. Commission only costs 1 pip per trade and slippage only a couple of pips more, max. So it's profitable.

    But the real profit will come from adding to positions on the trends as well as playing the chop strategies when the trend is flat and the counter trend strategies when the trend is active.

    Since the average winning trend lasts 3 months while the average loser only lasts 7 days, there's plenty of time to add to winning trades.

    The counter trend strategies, as you know, buy dips in long trends and sell rallies during short trends.

    It's glorious fun building this.

    It allows me to create simple strategies that I can interconnect and organize in any manner. Each one can serve as an input to one or more. Each can be set to totally different time frame of data. And the indicators inside each one can have a separate time frames from the others.

    The optimization system is finished. It takes less than 10 minutes to do 700 optimizations on 5 years of tick data.

    The analysis system (which is another pluggable component) reports the Profit Factor, average win, and other important stats.

    As part of that effort, it will also get a position sizing component to decide position size.

    The next step is to integrate it with the MB Trading interface so it can get live quotes and start trading live.

    The final step will be to have it run on multiple forex instruments simultaneously.

    After it has sufficient verifiable track record, I will seek O.P.M. to increase the capital for trading.

    Sincerely,
    Wayne
     
    #49     Jun 30, 2008
  10. Hey, I greatly appreciate all the suggestions on this thread. Most of them where extremely helpful.

    Sincerely,
    Wayne
     
    #50     Jun 30, 2008