Trend following on the short side

Discussion in 'Automated Trading' started by HuggieBear, Feb 13, 2011.

  1. I have spent a couple of years building and trading intermediate and longer term trend following systems. I've developed a few systems mostly focused on equity indexes as well as commodity futures, and I've had pretty good success.

    I've spend a tremendous amount of time trying to develop some decent mechanical short trend following systems but haven't been able to come up with much. INtegrating a system that trades short entries would be fantastic for smoothing out my overall equity curve and for reducing volatility, but the techniques i use on the long side don't seem to work very well.

    Just curious if anyone has had much success building longer term mechanical systems for shorting either indexes or commodities? I do have one system I am using on the short side in commodity futures. The system is pretty bad when looked at indiividually, but when incorporated into my overall strategy it does tend to reduce volatility a bit and provide a few percent alpha. However, it seems like so much more should be possible...
  2. Interesting post. I also tried in the past for several years to develop trend following system for the short side. One of the problems is that uptrends reverse suddenly and by the time you get your short signal you are already set for a correction and that hurts.

    I recently designed a crude oil short trading system using patterns derived by a program. I have posted some results before:

    I got a signal on 02/01 just when the reversal started and another on 02/07. Net profit close to $3 per contract, much better than trend following. Actually, most indicator based systems generated signals late last week and they are set for a good correction.
  3. Thanks for the info...that is an interesting analysis. Does PAL have a ton of patterns built into it that it automatically searches for?

    It is uncanny how frequently my short side trend following system sells the low only to be hit with a sharp reversal. I hate to say it but i think the right strategy might be to sell rallies in a confirmed downtrend....of course there are a few problems with that approach:

    1) autodetecting confirmed downtrends is harder than it sounds

    2) how to know the rally is not a full on reversal to the upside

    I may give this approach a try...i am not too optimistic, because selling rallies in downtrends mean you miss those great downtrends that don't have any rallies. But, its worth a shot...
  4. Isn't that in itself contradicting? You want to follow trends but when the trend is up, you want to short. What's the point?

    The only think I can think of is to short at the first warning sign and reap a few points with a profit target. If the trend resumes up, you have garnered a little bit from the short side and are still positioned long. If it goes down further, your long position will be stopped out anyway and you get a way with a smaller drawdown overall.

    Everything else wouldn't make sense.
  5. that's why systems can do better with human discretionary input.
  6. depending on the human, of course.

    Having tested so many different approaches, i just find it curious how difficult it is to find something that works on the short side. I suppose that is due to the natural tendencies of all things to rise in price.
  7. I don't think so from the program description. It is not based on a brute force search but on some other methodology. It is neither NN nor data mining.

    Things are easier than that I think. Selling rallies is too risky because when a rally takes place the short-term trend confirmation is no lonnger valid

    This is an example of the performance of a short-only pattern system from PAL (its previous version actually) during the collapse of 2008, from August to November. The maximum QQQQ net drop was about $20 or close to 40% loss in three months, a huge drop. The system with the patterns made $12 in 3 trades. That is good enough for me as far as catching short trend:


    I must admit I lost money from a long system I was using but overall the net result was positive. The short-only system worked perfectly.
  8. I am generally testing my short systems on between 10 and 30 years of data....over long periods like that, i can't find anything that works very well. On the other hand, it is relatively simple to find a working long system.

    Testing a short system during one of the steepest downtrends in history is bound to be pretty effective. The question is what did that same system do the 80% of the time the market is going up?

    You are right about selling rallies...i know thats not going to work.

    The problem with selling in the direction of a confirmed downtrend is that downtrends tend to be punctuated by sharp rallies. And it seems like those rallies occur invariably just after the downtrend gets confirmed.

    I actually did develop one system that did OK on the short side, but basically it had to have a VERY wide stop off the entry to deal with those snapback rallies. Not very desirable for me given the leverage at play in commodity futures...
  9. fjpenney


    Ed Seykota said that he found trying to make money by shorting to be about three times as difficult as making money with long trades.

    It comes as no surprise to me as an ETF trend trader that you find it difficult to develop a profitable algo for short trades.
  10. Do you trade long-only or short side as well? If short also, mechanical or discretionary?

    #10     Feb 13, 2011