NG collaboration development

Discussion in 'Journals' started by 40yotrader, Jul 11, 2006.

  1. bolter

    bolter

    Here's some thoughts on NG COT data that I put together lte last year. It might provide a few clues.

    Small Spec
    These guys are always net long. Since 1991 they have been net short only 1 week (Dec 2001). Their positions are very stable which suggests there is very little speculative retail money in this markets - ie: dentists do not have the scrotal fortitude to play this market. In the last 5 years their short holdings have remained constant but they have gradually added to their long positions. Possible impact of the emergence of long only commodity funds. Because of their long bias they make money during rallies and give it back during selloffs and sideways markets. Overall they don't do much better than break-even.

    Large Spec (Trading/Hedge Funds)
    These guys generally make money in this market. They take money off the small specs and commercials. They had a long bias up to 2001, but since then have definitely favoured the short side. They were net short during the recent explosive rally - got killed obviously. Their positions are alot more volatile than the small specs and commercials, in accordance with the market action.

    Commercial Hedgers
    These guys have a definite short bias. In fact whenever they get net long it is time to buy Nat Gas. Over time they lose money - they are hedging their forward requirements. Their positions have remained vitually unchanged for 3 years.

    Bottom line: Nat gas is a traders market. The regular explosive moves are not the result of a sudden increase in buying/selling interest - it is probably driven by interests in the pit.

    bolter
     
    #31     Jul 16, 2006
  2. m4a1

    m4a1

    here's another short only model. it uses 30 minute bars but does hold overnight. it also trades year round.
     
    #32     Jul 18, 2006
  3. bolter

    bolter

    m4a1,
    That looks like a decent model. Personally, I feel more comfortable with something that trades more frequently. You need a good deal of discipline to trade a model day in day out that might only generate one or two signals a month.

    Also, I much prefer models that trade from both sides of the market. Results from trading just one side would ordinarily constitute a selection bias (ex post) in your testing.

    However, in this case it actually highlights some important characteristics. And that is, like many commodities, the mean price for NG over time is reasonably stable (ie: does not exhibit any drift unliek say equities), in a low inflation environment, and following the inevitable supply (vesus demand) shocks price does mean revert. That's definitely an inefficiency that can be exploited.

    What happened to 40yo - I guess he got bored with his own thread?

    bolter
     
    #33     Jul 18, 2006
  4. m4a1

    m4a1

    yeah the number of trades is too low, bummer. this is a typical problem for me. increase trades and the profit factor goes down. increase the profit factor and the number of trades go down.

    i went after a short only model because i figured since it has been such a bull market, finding a long only strategy won't "prove" anything.

    how many trades does your model generate per year and what's your profit factor?
     
    #34     Jul 19, 2006
  5. bolter

    bolter

    m4a1,
    Bear in mind that my system is end-of-day so the trade frequency will be much less than an intra-day system.

    Having said it generates about 1 trade per month, but it is in the market most of the time. Profit Factor is 2.4. Stats are attached for the entire 16 year history of NG.

    Not suprisingly the short side performs better than the long. What is unusual is the 52% win rate - which is excellent for a long term trend following model.

    <a href="http://tinypic.com"><img src="http://i2.tinypic.com/20g1ez7.png" border="0" alt="Image and video hosting by TinyPic"></a>

    All the best,

    bolter
     
    #35     Jul 19, 2006
  6. m4a1

    m4a1

    bolter, i'm not sure if i understand what the max drawdown numbers in your report is saying. why is the combined max drawdown ($-38,486) lower than the max drawdown from the individual sides ($-56,475)?

    is your profit factor and win rate stable year to year?
     
    #36     Jul 19, 2006
  7. bolter

    bolter

    m4a1,
    Of course the drawdown will be highre/lower for one side - if you think about it.

    Are the stats stable over time - yes for the most part. Go back to my original equity curve - it is fairly stable.

    regards
     
    #37     Jul 19, 2006
  8. m4a1

    m4a1

    oh yeah, doh! i was thinking of tradestation's "max drawdown" field which actually means something else.

    have you used tradestation before? how does wealthlab compare?

    i think 40 decided nobody can help him.


     
    #38     Jul 19, 2006
  9. I've talked to a broker that specializes in NYMEX. He told me to expect slippage of $300 r/t and not to plan on doing more than 10 contracts at a time. In effect, he told me I was wasting my time looking at NG. I've decided to abandon this market and go back and toil on other projects in obscurity. I turned off my PM's because I'm tired of all the pimping on my original ES system.

    Good trading to all

    44yotrader
     
    #39     Jul 19, 2006
  10. m4a1

    m4a1

    here's the short only model with $500 slippage per round turn trading 1 contract. added $200 just to give it more cushion.
     
    #40     Jul 20, 2006