Mechanical System for Trading the E-minis

Discussion in 'Strategy Development' started by Ardit13, Sep 29, 2005.

  1. Ardit13


    Here's a mechanical trading system for the e-minis that's posted on a free site.

    How do I know if this is a good trading system or not? What do system traders look for in these performance summaries?

    What's the most important piece of data? Is it net profit, profit factor, average trade net profit, percent profitable, or something else?

    What does the profit factor mean?

    What does it mean if a system is over-optimized?

    I have a lot more questions, but I'll leave it at that for now.
  2. Ardit13


    I see a lot of people viewing this thread, but no one leaving any explanations. Maybe the TradeStation summaries will help.

    I can't figure out how to post the TradeStation performance result graphics (at least no graphic is showing up in the preview). You can review the performance results for the system here:

    I don't know if these are standard TradeStation system summaries.

    How do I read the summaries to tell me if a system is any good?
  3. I don't think there's any ONE thing that people look for. Alot of it depends on how well a system fits in to your portfolio. And that is different for everyone.

    My thoughts on the system you asked about...

    The profit factor is a little low. I like to see my systems around 2. It doesn't mean a system isn't tradable if it is below though.

    Percent Profitable is nice. I like to see over 50%, again 40% wouldn't neccessarily mean it is a bad system.

    Average trade is low on the s&p and nasdaq. I'm assuming there are no commissions and slippage calculated in the reports, so after you account for that you aren't left with much. That makes the system untradable on those markets IMO.

    Drawdown is also a little high for me. I like to keep a drawdown/netprofit ratio of .1.

    That's about what I draw from a quick glance at those reports. Hope this helps!
  4. Markus



    since you "insist" on an answer, here are my 2 cents:

    The first figure to look for is the Net Profit. Obviously you want your system to generate profits.

    The next figure you want to look at is the Average Profit per Trade. Make sure this number is greater than slippage and commissions, and that it makes your trading worthwhile. Trading is all about risk and reward, and you want to make sure you get a decent reward for your risk.

    Take a look at the Profit Factor (Gross Profit / Gross Loss). This will tell you how many dollars you are likely to win for every dollar you lose. The higher the profit factor the better the system. A system should have a profit factor of 1.5 or more, but watch out when you see profit factors above 3.0, because it might be that you over-optimized the system.

    Here are some more characteristics you might want to consider besides the net profit of a system:

    Winning percentage
    Many profitable trading systems achieve a nice net profit with a rather small winning percentage, sometimes even below 30%. These systems follow the principle “Cut your losses short and let your profits run”. However, YOU need to decide whether you can stand 7 losers and only 3 winners in 10 trades. If you want to be “right” most of the time, then you should pick a system with a high winning percentage.

    Number of Trades per Month
    Do you need daily action? If you want to see something happening every day, then you should pick a trading system with a high number of trades per month. Many profitable trading systems generate only 2-3 trades per month, but if you are not patient enough to wait for it, then you should select a system with a higher trading frequency.

    Average Time in Trade
    Some people get really nervous when they are in a trade. I have heard of people who can’t even sleep at night when they have an open position. If that’s you, then you should make sure that the average time in a trade is as short as possible. You might want to choose a system that does not hold any positions overnight.

    Maximum Drawdown
    A famous trader once said: “If you want your system to double or triple your account, you should expect a drawdown of up to 30% on your way to trading riches.” Not every trader can stand a 30% drawdown. Look at the maximum drawdown the system produced so far, and double it. If you can stand this drawdown, then you found the right system. Why doubling? Remember: your worst drawdown is always ahead of you.

    Most consecutive losses
    The amount of most consecutive losses has a huge impact on your trading, especially when you are using certain types of money management techniques. Five or six consecutive losses can cause you a lot of trouble when using an aggressive money management.
    In addition this number will help you to determine whether you have enough discipline to trade the system: Will you still trade the system after you have experienced 10 losses in a row? It’s not unusual for a profitable trading system to have 10-12 losses in a row.

    Now you know.... :)

  5. Ardit13


    I appreciate the effort you put into your responses. I haven't found better explanations searching the web.

    I like the idea of doubling the maximum drawdown. Simple, yet original (at least to me).

    When would you know if a system has been over optimized? The profit factor is too high? Winner percentage too high? Not enough trades? Is there something in a TradeStation summary that would instantly lead me to believe a system is over optimized?
  6. extremely high profit factor, not a large enough sample of trades, if you look on the settings tab you can see how many parameters were "possibly" optimized. Just because they are there doesn't mean they were optimized.
  7. Markus


    1. Look at the rules

    You can eliminate almost all losers by adding enough rules. Simple example: If you see that on Tuesdays you had more losers than on the other weekdays, you might be tempted to add a “filter” that prevents your system from entering trades on Tuesdays. Next you find that in January you had much worse results than in other months, so you add a filter that enters trades only from February – December. You add more and more filters to avoid losses, and eventually you end up with a trading rule that I saw recently:

    IF FVE > -1 And Regression Slope (Close , 35) / Close.35 * 100 > -.35 And Regression Slope (Close , 35) / Close.35 * 100 < .4 And Regression Slope (Close , 70) / Close.70 * 100 > -.4 And Regression Slope (Close , 70) / Close.70 * 100 < .4 And Regression Slope (Close , 170) / Close.170 * 100 > -.2 And MACD Diff (Close , 12 , 26 , 9) > -.003 And Not Tuesday And Not DayOfMonth = 12 and not Month = August and Time > 9:30 ...

    Though you eliminated all possibilities of losing (in the past) and this trading system is now producing fantastic profits, it’s very unlikely that it will continue to do so when it hits reality.

    2. Test it with variations of the parameters

    If the system produces similar results when you vary the original parameters (e.g. bars to be used for a moving average) by 15%, then you have a quite robust trading system.

    3. Test it on multiple markets

    Especially related markets: If you designed a system for the e-mini S&P then it should perform well on other indices, too. Run the system on different markets to make sure that you haven’t optimized the system for a single market.

    Hope that helps.

  8. Ardit13


    When you write "bars to be used for a moving average," what do you mean by that? If you are using a 100-period moving average, see what happens with a 115-period moving average and an 85-period moving average?

    Here's a product that others on Elite Trader have slammed. It has 85 trades over 15 years, with a profit factor of 9.45 and percent profitable of 86%. Is this a good example of over optimized? Is there anything specific that tells you this is over optimized?
  9. Markus


    Yes, exactly.
    If the system uses a 100-period moving average then you should test it with
    85, 90, 95, 100, 105, 110 and 115.

    It should still be profitable and variances between the results should be rather low.

    Take a look at the picture below. The underlying system uses Bollinger Bands with a 34 bar setting. The table shows performance reports for different settings. As you can see, the results vary, but not very dramatically. The system seems to be robust.

    If in this example the system produced profits with a setting of 34, but losses with any other setting, then the system is overoptimized.

  10. Ardit13


    Of course you were kind of enough to show us the results that were NOT a 34-period moving average. But in most cases, we don't get that extra information. We only get to see the results of the actual system (e.g., the 34-period MA system). So looking at the TradingStation output of the two sample systems above (or any other system output), is there something specific in those numbers that would tell you if the system is over-optimized or not? Is there something an experienced system trader (or tester) will see in the output and say, "Ah, what a load of crap. It's over-optimized."?
    #10     Sep 29, 2005