backtesting trading systerms

Discussion in 'Strategy Development' started by dtrader98, May 31, 2007.

  1. I posted this in automated trading but figured this might be a better place.
    ------------------------------------------------
    For the old timers here. Are there any good threads that simply look at long term mechanical systems and compare metrics
    (CAGR, Drawdowns, num. trades, etc.)

    I often see a lot of systems shown, and have done a lot of backtesting myself, however, i rarely see systems knocking out simple buy/hold approach. Not only that, but rarely do the systems account for short term tax vs. long term cap gain.

    I'm thinking long term systems --at least 20-30 yrs worth of data. Most of the systems (Dr. Elder, Tharp) for example advocate 3.5 ATR stops, but when i run long term data, it seems like 7-8ATR is MUCH better in performance.

    Even Curtis Faith showed better performance ignoring any type of calculated stops and just using time based exits instead.

    A perfect example would be if someone could show me a prior mechanical system posted, that beat a major index B/H over say a 30 yr. stretch.

    BTW. I'm not seeking a backtesting platform, just a comparison of mechanical trading systems.

    Thanks.
     
  2. if you don't mind, what kind of software do u use for backtesting?
     
  3. Check out the tradingblox forum (http://www.tradingblox.com) , there are a couple threads there of people comparing different long term trading systems and their risk adjusted returns.
     
  4. Prevail

    Prevail Guest

    I'm not aware of a thread with the info you seek but maybe search for it.

    finding a systm profitable for 30 years to out perform b&h is tough.

    pm me if you want to see one I watch which has done so, but has large dd's.
     
  5. Curtis' site was pretty useful. Thanks Makloda. They are discussing mechanical systems the way that they should be approached IMO. And it pretty much cooberates what i was seeing; 3.5 ATR on long term systems (like ma cross) is much worse then 7-8ATR (or essentially no ATR stop as ATR>8 pretty much equals the original system with no ATR stops).

    However, I am finding a few systems that do beat B/H long term, but they do not use simple indexes, they are heavily diversified in uncorrelated markets, which steps up the complexity over the systems proposed in most trading books to date IMHO.
    Way of the Turtle touched on this slightly.
     
  6. Is this what you are looking for?

    This system purchases Altria Group, Inc. stock symbol MO when the value of the 70 day exponential moving average is greater than the value of the 130 day exponential moving average. Position size is 0.03 * equity / (10 * the 20 day average true range). Stock is sold when the value of the 70 day exponential moving average is less than the value of the 130 day exponential moving average. Commission is $ 10 / transaction. Slippage is estimated at 50 % skid. This model uses 34.41 years of daily historic price data beginning 2 January 1973 ending 25 May 2007. Trading is long positions only.

    ===

    Number of trades 9
    Total profit $ 1107745
    Profit after subtracting $ 10.00 commission, slippage per transaction: $ 1107565
    Heat is 3.00 per cent of equity.
    Greatest drawdown is 0.0465 (4.65 per cent).
    Cumulative Annual Growth Rate (CAGR) is 32.19 per cent.
    CAGR / Drawdown is 6.92
    Instantaneously Compounding Annual Growth Rate (ICAGR) is 7.24 per cent.
    Annually Compounding Annual Growth Rate (ACAGR) is 7.51 per cent.
    Information Ratio is 0.67
    Initial capital is $ 100000

    ===

    Buy And Hold Method Investment Analyzer
    Years of data 34.41

    Cumulative Annual Growth Rate (CAGR) is 854.16 per cent
    Greatest drawdown is 64.88 per cent.
    CAGR / greatest drawdown = 13.16

    Instantaneously Compounding Annual Growth Rate (ICAGR) is 16.53 per cent.
    Annually Compounding Annual Growth Rate (ACAGR) is 17.97 per cent.

    ===

    I recall many instances of the buy and hold model showing greater profit than a trend following mechanical system. I remember the trend following mechanical systems usually show lesser drawdowns than buy and hold models.

    In this example buy and hold shows a greatest drawdown of about 65 %. The trend following system shows greatest drawdown of about 5 %.

    I choose a mechanical trading system that shows a history of moderate growth rate and lesser drawdown. Other traders might choose buy and hold, accepting the history of greater volatility, perhaps seeking greater returns. Ultimately the choice of trading system is a personal decision.
     
  7. That's a great example.

    I was looking at covel's book, with a bunch of curves showing at least double the performance of b/h. The methods were proprietary and didn't indicate impact of taxes. In reality you have to take out tax on every short term gain, which about cuts those curves in half, bringing them to about the same as b/h with more work.

    However, there are some systems he divulges in the back. One was buy/short nas 100 using 100 day donchian ch breakouts. The portfolio curve was up like 18X!!! from 90-2004. Whereas the index itself during that period went up only ~7X. It also used a 4TR stop.

    Hard to believe that kind of performance, although i haven't run that specific system.

    That's one of the reasons I was curious to see if others see similar types of performance on simple non-leveraged, self correlated indices.
     
  8. Great post. I learned it is a misconception to use a ruleset/system solely to "outperform" returns. A rule set can also be used to simply lower volatility while accepting lower overall returns, but thus giving you the chance to leverage it up by 100% and still achieve overall higher risk adjusted returns with lower draw downs than buy and hold.