ES SMA Cross Over Back Test Results - Too Good to be True?

Discussion in 'Index Futures' started by xmlschema, Aug 18, 2007.

  1. I ran an ES back test on a SMA Cross Over strategy. Time period is from 4/1/2007 - 8/17/2007. I know this is very narrow but it's pretty much all I had to work with.

    Looks like the system made over $10,000 (exluding commission charges) trading 1 contract. Largest drawdown was just at 2% for no more than a day or so.

    Is this too good to be true? Am I missing something? I'm attaching the results for anyone to disect.

  2. Are the results highly dependent on the periods you're using for your SMA crossover?...If so, your results may be curve-fitted.
  3. While I do not use an MA as a basis for entering any of my trades there is a reasonably effective rule that can often produce consistently decent trades on the ES

    Set up your charts using a 5, 15 and 45 minute interval.

    Set up your 2 simple moving averages using 7 and 34 bar lengths.

    Color the 7 bar green and the 34 bar yellow.

    Rule 1
    Never go long if the 5 minute bar is less than the yellow line.
    Never go short if the 5 minute bar is greater than the yellow line.

    Go long on the first bar greater than a green line when it is greater than the yellow line.

    Go short on the first bar less than the green line when it is less than the yellow line.

    Rule 2
    Watch the spoos bounce off a 34 bar MA when the interval is set to 15 or 45.

    A better approach is to learn how to really identify the trends.

    If you have a cross over system programmed, I'd love to see the results using Rule # 1.
  4. During the period you tested, everybody was using the following trading algorithm:

    10 Start
    20 Buy
    30 Hold
    40 Go to 20
    50 End

    I'll bet that a lot of shit that ordinarily doesn't test out worked.
  5. I don't see why they would be dependent on the period I used (April - August). Those are the months I had historical data for.

    Would there be any reason why SMA Cross-Overs would be period dependent?
  6. if you go to the ES journal, you can see some pics, of similar systems.

    but moving averages are very slow in response time, and based on market conditions(ie choppy), you will get a lot of false crossovers and reversions, which ends up making the system buy at minute to minute peaks and selling valleys. (the thing you want to avoid).

    Moving averages work in markets that are trendy in nature. And then the next question would be, what markets are trendy and when are they trendy with minimal reversions. And then this gives you clues to the filters to setup.

    ES is unique derivative, and in it the participants or whales who are in that market participating, are making money using other algorithms, the moving average algorithm takes advantage of the their algorithm, there will come a time, when enough people are doing it that it wont work anymore. Meaning the ES price pattern will be more choppy, and the filters only get triggered a few times in a given year.

    In FX if you watch price behavior, on a day to day basis, its extremely chaotic, but during times of crisis such as the past few days when the carry trade was unwinding, the price patterns were literally throwing money at some of these algorithms.

    The latest thing are trendlines and channels, and the price pattern signals don't lag as much as moving averages. Jack Hersheys followers will attest to it.
  7. Pay attention to what hypostomus is saying here.

    During this time period we have experienced volatility the likes of which hasn't been seen on such a consistent basis since ... oh maybe the 2002 -2003 time period.

    While your MA crossover MAY work during other market periods, the only way you will know for sure is to test it out.

    You need to invest in about four years of data, if not more ... four months of (especially the past four) just won't get it done.


  8. I am referring to the periods of the SMA's you are using, ie your SMA parameters...for example: if you are using a 10-bar moving average crossing a 20-bar moving average, then i am referring to those "10" and "20" numbers. I am not referring to the historical time period you used in your test, which is April to August
  9. Good point. I guess that was what giggollo was alluding to.
  10. What about exits and stop losses?

    Also Re: rule 1, what about when price is below the yellow MA but above the green MA and it's rising, as soon as the green MA crosses over the yellow one, do you enter? Technically, all the conditions will have been met then (price above yellow, price closed above green)?

    Otherwise I see how it could work for buying pullbacks and stuff.

    I'm assuming you meant enter on the close of a bar over the MA.
    #10     Oct 31, 2008