System on/off switch

Discussion in 'Strategy Development' started by thetrendfollowe, Jan 20, 2008.

  1. Hi everyone,

    In October i just started trading a system live. This system trades equities long-only, and its strategy is long term trend following.

    The system is designed to take advantage of the long term historical bullish bias that the stockmarket has shown.

    Due to the recent activity in the stockmarket, and with the system close to reaching its maximum drawdown, I am thinking of adding a system on/off switch, also known as an Index filter. The purpose of this filter is to switch the system "Off" during a prolonged downtrend.

    Now during "corrections", I suspect such a switch won't do much to improve results as by the time I get back in to the market, I would have missed the aggressive rebound, so the end result is probably no different than if I had kept the system working throughout the correction period.

    BUT that's okay, as long as the switch keeps me out of a prolonged downtrend, of say, 6 months plus, then I suspect over the long term, the result, compared to the standard system, will be comparable profits, with less pain (drawdown).

    Though I haven't tested any ideas at this stage, what Im thinking could work, is to turn the system off if last weeks close was lower than the previous weeks close (The system is based on weekly bars). And turn it back on, when the index makes a new 52-week high. This would certainly keep me out of a prolonged bearmarket, yet get me back in without missing out on too much action.

    Does anybody employ an index filter/system switch into their long term trend following system?

    I would appreciate any thoughts or suggestions here.

    Thank you,

  2. If you began trading the system during October 2007 then I do not think enough time has elapsed. I observe that with a trend following system losses come first (because losses are taken quickly) and profits come much later (because profitable positions might be held for years). It may take a year or two for the system to accumulate a few big trending securities in the portfolio.

    You might consider adding the Standard And Poors index tracking stock SPY to your portfolio. My tests show SPY sometimes shows significant trends. Addition of SPY to the portfolio automatically increases investment when index values are increasing and decreases investment when index values are decreasing.
  3. Hi Hook N. Sinker,

    Thanks for your response.

    Yes you are correct, as average losing trade is about 2 months, and the average winning trade goes for 6-7 months, the system needs time to work.

    You cannot test a longterm system on a short timeframe and expect results. The system probably needs 2 years to show results. In every backtest, the first year was almost always (slightly) negative. That's not the issue here.

    The point im trying to make is that, as the system was only tested from 1998-2007, it has never been exposed to a prolonged bearmarket ie. worse than 2000-2002.

    So if we do enter a prolonged bearmarket, I do not want to trade the system as it is being traded outside conditions in which it was tested (which means i can no longer expect to get results similar to those found in backtesting). So I want a switch to turn the system off, and then turn back on once the market conditions are better.

    Of course I will first backtest any idea I come up with over the 2000-2002 bearmarket and over the whole historical data that I have to see if it made a difference. If it does, then i will implement it.
  4. MGJ


    One component of an analysis might be a historical backtest, comparing the system WITH the on/off switch, against the system WITHOUT the switch. You could plot the two equity curves on the same graph and watch as they move together or move apart
  5. pneuma


    I have an index switch in some of my trend following systems. It simply tracks in index of the stocks I am trading and when the index falls below a simple moving average the system doesn't take any new positions. For example, if i was trading the S&P500 components it would track the S&P500 and not take any trades if the S&P was below say a 50 day moving average. I have found that you can also use an advance/decline line instead of the index.

    Performance wise, I have found that without this filter the equity curve was up and down; with it i have found the equity to be flat and up - more like steps than peaks and troughs. It certainly improved performance for me.

  6. Perhaps you could try a monthly(or weekly) parabolic stop&reverse, either over the individual components, or an index filter as a guideline, should work ok-dont go long when its short.