When your system overrides work...

Discussion in 'Automated Trading' started by Cutten, Jul 4, 2009.

  1. Cutten


    As a diversifier I run a couple of trading systems with a small portion of my capital. I have a slight problem in that occasionally I feel a strong urge to override them because I see they are about to do something I think is wrong. Usually it is when I see them going one way when I have a strong conviction that the market is going the other.

    So far I've tried leaving them alone, and tried overriding them. When I have a strong view and don't override them, I usually regret it. For example in May 08 I got system buy signals in stocks, I overrode it because I was ultra bearish. In August 08 the system was fully long commodities, and I got seriously spooked by the huge crude oil selloff from 147 to below 130, I exited it and saved a lot of money again. It was long bonds in Q4 2008, and earlier this year I trimmed the position by half as the bonds just kept tanking (I only trimmed half because I was - incorrectly - mildly bullish on bonds). I lost half as much as it would have normally.

    The worst example was after the swine flu scare. The system was already short lean hogs, and added on the breakdown on the swine flu news. The market puked over 15% in a few days, and I felt it might be a bit oversold. I didn't override because my conviction was not huge, and the main trend was still strong. The market then rallied back to close the gap it made on the swine flu news. Now, as it was rallying, I felt with about 90% certainty that it would probably close the gap and stop out all those who had shorted on the news, then turn back down after squeezing out the shorts, and make new lows. However, my system had its trailing stops at exactly that price point. I decided to keep 1/4 of the short position on and move the stops higher than the gap, and let the other 3/4 position trade according to the system (I was trying to avoid overrides as it kinda defeats the point of systems trading).

    Anyway, sure enough the market closed the gap, stopped out the system literally within about 4 or 5 ticks of the high off the move, then immediately puked 20% to new lows in the next couple of weeks. My discretionary override did really well but I basically missed out on an extra 20k profit because the system exited at the worst possible time.

    I don't get the urge to override normally. My main issue I think is that I find it psychologically bothersome to watch the system do something "dumb" and not do anything about it.

    So, I see a few alternatives:

    1) Stick to the systems. They are net profitable and make acceptable returns with acceptable risk, and are almost totally uncorrelated with my normal trading results.

    2) Use the systems but override them when I have strong conviction.

    3) Throw away the systems and trade pure discretionary.

    4) Redesign the old systems to incorporate the discretionary input that seems to be making money, if this is possible (I'm not sure it is, as my discretion is subjective feel rather than objectively measurable).

    Any thoughts?
  2. FWIW I trade in a semi-discretionary manner now, automating only elements of the strategy.

    The reason I do this is that I haven't been able to figure out to get an automated version of what I do to qualify "context" properly. I like the automation that I have because it frees my time and attention.

    What you appear to be doing is a lighter version of the same. You automate but sometimes context breaks strongly from the "normal" conditions that your automation is designed to operate under. To me that sounds ok but in your situation I might add a qualification:

    Try to define normal and what might be an acceptable break from normal such that you should override the system.

    Such a qualification may not be necessary, however, if the large percentage of your overrides are working for you.
  3. I had the same issues to figure out and I took the route of trading the system as it was designed until I felt I had a reasonable sample size to make a decision off of.

    I ended up trading the system as it was because on the whole I did worse that the system over the sample of trades I investigated.

    It sounds like you're the opposite of me. IMO I dont think the term 'system' means that you don't tweak/modify/override it sometimes if you perform better than the system on the whole.

    I think a lot of the 'don't override system signals' logic is good for those that perform worse than the system but it sounds like thats not the case for you. I use automated systems to handle some parts of a strategy & handle others manually. That seems to work out best for me.

  4. Actually depends on the nature of your systems. I am developing my own at the moment (first one is NOW in forward testing). Whenever I see the system do something stupid of have objections...
    ...i think of adding a filter to the system.

    Note tha tthis is not necessarily "curve fitting". I dont really change the parameters of the system. I just look for areas where I say "ok, that is just outside of the parameters the system should actually even try working under".

    Example: My system is a counter trend system. Works well in sideways markets, even light trends. Totally bullocks, concept wise, in runaway markets. Yesterday in the YM rally, it made 3 attempts to make money by SHORTING into it - 2 total blows, one a light profit.

    Well, looking at the chart I realize - the graph goes straight up, this is not where the market is intended to work. Lets add a filter "Strong Trending" and NEVER open position against it. Point. This is like the third filter I added, and they all go around "filtering out where the system should not even try working".

    You could do the same.

    Otherwise you have to montiro and trust the system. Put it on a separate (sub) account, and trade your own ;)