What do you mean by this ?
Discretionary Versus Mechanical System Trading
- Thread starter Murray Ruggiero
- Start date
What do you mean by this ?
Exactly what I wrote. I tried to backtest a few trend following systems and all of them gave poor results on indices like SPX, DIA, although they seem profitable on several equities.
I also found that others, more experienced system traders are of the same opinion, i.e. well respected Chuck LeBeau:
Are you of a different opinion? Please share.
Murray Ruggiero
Sponsor
If we are talking about maximizing performance of a trading strategy for the stock indexes I would agree trend following does not work. If we are talking about just outperforming the market returns then strategies like the 4% rules on larger basket indexes like the Value line and the Russell 2000 can outperform buy and hold.
For the trading community you need to use counter trend methods to trade the indexes, for the investor, some of the trend following methodologies can cut your volatility and risk over buy and hold.
For the trading community you need to use counter trend methods to trade the indexes, for the investor, some of the trend following methodologies can cut your volatility and risk over buy and hold.
Dude. Couldn't say it better myself. There is truth here. Infer what you will.
Quote from cnms2:
Mind will help you some times will let you down other times. Statistics is objective, but can be brought down by mind: both through data fed in, and through result interpretation. You have to use them properly and cautiously.
Well, there are certain advantages to mechanizing a well developed rule set;
1. The program always follows the rules
2. The program doesn't "sleep in" or "miss signals"
3, The program can trade multiple markets concurrently
these advantages are undeniable, and as a result, when the program is well thought out and an edge exists, programs can't be beat.
Unfortunately, these conditions (well developed, robust, with a verifiable edge) are almost never found all together. Most programmers do not understand or know how to find a true statistical edge. Instead they often develop a system based on historical tendency. That is why almost all systems do well at times, and poorly at other times. Also (of course) it is true that some statistical edges are unstable.
While good programming offers the advantages that I have suggested, one has to realize that most programs will not adapt or change to a changing market condition. Instead the programs simply stop being profitable. In contrast, a good discretionary trader can adapt on the fly, and take advantage of changes in a market. Also a good discretionary trader can do other things that a program (usually) cannot, including modifying position size, stop loss, and (when necessary) standing aside when market conditions are adverse. Admittedly, these are limitations of the programmer not the technology.
The comment that trend following does not work for indexes was offered without much thought, and unfortunately it was mis-interpreted by the poster. Mr. LeBeau was commenting on a specific type of trend following known as the "Turtle System". Although many of you throw around the term "Turtle System", few of you really know in detail, what the system requirements were. In point of fact, I use trend following as an important part of my own intraday system, and as of yesterday, it was still working pretty well. I do not use the "Turtle System".
Good Luck Meatpuppets
Steve
1. The program always follows the rules
2. The program doesn't "sleep in" or "miss signals"
3, The program can trade multiple markets concurrently
these advantages are undeniable, and as a result, when the program is well thought out and an edge exists, programs can't be beat.
Unfortunately, these conditions (well developed, robust, with a verifiable edge) are almost never found all together. Most programmers do not understand or know how to find a true statistical edge. Instead they often develop a system based on historical tendency. That is why almost all systems do well at times, and poorly at other times. Also (of course) it is true that some statistical edges are unstable.
While good programming offers the advantages that I have suggested, one has to realize that most programs will not adapt or change to a changing market condition. Instead the programs simply stop being profitable. In contrast, a good discretionary trader can adapt on the fly, and take advantage of changes in a market. Also a good discretionary trader can do other things that a program (usually) cannot, including modifying position size, stop loss, and (when necessary) standing aside when market conditions are adverse. Admittedly, these are limitations of the programmer not the technology.
The comment that trend following does not work for indexes was offered without much thought, and unfortunately it was mis-interpreted by the poster. Mr. LeBeau was commenting on a specific type of trend following known as the "Turtle System". Although many of you throw around the term "Turtle System", few of you really know in detail, what the system requirements were. In point of fact, I use trend following as an important part of my own intraday system, and as of yesterday, it was still working pretty well. I do not use the "Turtle System".
Good Luck Meatpuppets
Steve
Agreed. Several years ago when I first got the idea of developing my automated system I didn't know much about markets and my first instinct was to look for patterns and rules that I could code into logic. After much experimentation and reasoning I decided this was not the right approach.
After massive amounts of research into non-linear problem solving, statistics, probability, optimization, etc I've finally developed a 'self-tuning' system which uses a form of 'online' non-linear optimization so I don't have to really worry about my system ceasing to work one day.
For my system to stop working there would have to be some very very large structural change in the markets, which is possible still, but not as likely.
After massive amounts of research into non-linear problem solving, statistics, probability, optimization, etc I've finally developed a 'self-tuning' system which uses a form of 'online' non-linear optimization so I don't have to really worry about my system ceasing to work one day.
For my system to stop working there would have to be some very very large structural change in the markets, which is possible still, but not as likely.