Suggestions needed about amount of data needed.

Discussion in 'Data Sets and Feeds' started by blox87, Aug 7, 2010.

  1. blox87

    blox87 Guest

    I have a system that looks great over the past two years... Is this equity curve possible in real life or is it just a spoof in the backtesting? I've checked every single facet making sure that future data isn't being leaked . This is my most successful system i've built in two years and wondering if any others have achieved a similar or better equity curve. ( probably). Could anyone help me out with a suggestion or two through PM . This is using 1 contract. Similar curves in the ES, NQ, and YM. If I use it in forex it is bunk. I need ten years worth of data for ninja trader for me to feel comfortable using this . Where do I find that?

    Thanks in advanced
  2. blox87

    blox87 Guest

  3. I gather that is a 5-minute timeframe.

    Your avg. win to avg. loss ratio is very low for your win rate, resulting in a marginal profit factor. This may indicate that your results could have been obtained by chance alone. I would like to see a profit factor greater than 3 for this low win rate to be sure there is no luck or optimization involved.
  4. You can get about 8 years from
  5. Your results don't include the effect of commissions. What are the results when you do?

    Also, could you post the results in "%" (rather than "Currency") terms?

  6. blox87

    blox87 Guest

    $10,000 dollars in commissions, $72,000 profit including commisions from 1-1-06 to present using 1 contract.

    Max draw down 10% ($4437)total profit 314% ($72000 ) including commissions

    I'm buying historical data today for the past 10 years to see how this baby does. It does even better in the NQ does very well, in es, and tf also. I'm pretty sure I have a clearly defined edge. Hopefully i'm not fooled by randomness. I find if something is too good to be true it usually is but we shall see how this 10 year test goes for the final decision of whether to commit any capital to it or not. Wish me luck that it wasn't just luck! lol
  7. You don't need all that back data.

    What works now essentially works now. Why care about 2003?

    But the only thing is are you sure somehow you are not getting preferential fills etc?

    Was that drawdown max loss? Was it max peak to valley?

    What is your MAE over this whole time?
  8. I am confused, it sounds like all you have done is back test.

    Want to know if you have a great system - execute it with real money. If your results are not within 5% - 10% of your test, then you should consider a new app for system design and testing.

    Markets change, a system that has been successful for the past 6-months, 2-years or whatever - does not guarantee future success.

    As far as profitability, I won't use a system unless it returns at least 4% average a day - 12% daily return is optimal - anything higher is usually indicative of over optimization.

    This has been my experience.
  9. blox87

    blox87 Guest

    MAE $93 bucks on 1 es contract

    MFA $203

    The max drawdown is from peak to valley. The worst actual drawdown for eaquity was right at the beggining . I was in the red $275 for a bit but after that it was nothing but profits.

    I guess the real test will be live and whether fills ect really screw the whole thing up. I would like to get a higher profit factor I guess to feel %100 percent comfy with the system but who ever said trading is supposed to be nice n comfy right.

    Ill try it on the nq live . My max drawdown from peak to valley on that one was $2200 using 1 contract so I'd be willing to risk 5 grand to see if it works live. If the max DD hits 5 grand then I stop using the system.

    I want that 10 year data so I can get a feeling of what kind of draw downs will be expected. Thanks for the opinions so far.

    To the gentleman that gets 4% a day ... That's awesome and I hope someday I can get to that point . Seems like it would be impossible but I bet it is possible to a very very fortunate few.
  10. Your back testing should include predicted slip on any order that is filled.

    The impact of slip on a trade can be quite significant - especially if you make a lot of trades.

    Also, you have tested on 6-years of data, that is a lot, I test on 3 contract periods at a time (I currently trade the S&P 500 e-mini)

    One word of caution - markets really do change when you least expect them to do so.

    I have found that if you over optimize, your system may become brittle (not robust). I generally go with less than optimized settings that are more robust to changing market dynamics.

    Anyway - run the damn thing with real buck$$ - it is the only way you are going to know.
    #10     Aug 9, 2010