I remember reading something out of a book, maybe it was The Psychology of Trading, and the author said something to the effect of, "Don't try to make a mechanical system. Back testing is useless. The persuit of mechanical strategies is the result of being fed for doing nothing while you were a baby. Now you are trying to recreate being fed for nothing with your trading system. No system can analyze the complexities of the market like the human mind." Again, I was paraphrasing, and I may be remembering it wrong, but does the author have a point? Is system trading fruitless? I have no doubt that people can find profitable systems, but will those systems last? Or are systems simply a crutch for those without discipline? I sometimes wonder if it is possible to quantify everything that goes on in a traders head. Maybe there is some utility in system development because it separates genuine trading logic from uselss "gut feelings." However, it is also possible that trading sytems could prematurely dismiss not-easily quantifiable, yet legitimate "gut-feelings." Or perhaps the best way to use a trading system is to objectively define your strategy, so that you will be more disciplined. But if you then use that system robotically, maybe you are throwing away all that your mind is capable of that a computer is not, namely adaptability. Or maybe I'm rationalizing cus I suck at programing.
Whoever was the author of that quote is naive at best, and an idiot at worst. I have no doubt that the most profitable traders in the country are using automated trading. I am talking about traders and trading institutions making tens and hundreds of millions per year. The only shortcoming of automation is that it can be difficult and time consuming. However, for those that have done it, they will never look back. In five years, I think 90%+ of profitable traders will be using automated systems, if not sooner. I've been doing fully automated daytrading since 1998, and would never have experience a fraction of my trading success without it. Some can be very profitable using gut feel and fully 'discretionary' trading, and my hat's off to them. For me, automation is the only way to go. My two cents... Best of luck, -Eric
(...but does the author have a point?) Yes, to promote discretionary trading (Is system trading fruitless?) No, unless my system's results of 100% from 1/1/02 are not a fruit (...but will those systems last?) Depends on the system. If its good ,it will out last most peoples dedication to trade by 90% - Just ask your broker how many of his clients are profitable (Or are systems simply a crutch for those without discipline?) To trade a system requires the utmost discipline (I sometimes wonder if it is possible to quantify everything that goes on in a traders head.) A billion dollar number crunching Cray supercomputer has 100,000 times less connections than our brains have and processes data about 100,000 times slower - so NO!, it is not possible (...maybe you are throwing away all that your mind is capable of that a computer is not, namely adaptability) The best systems are constantly adapting to the current market enviroment.
Yes, mechanical systems work. No they don't work forever. And extended backtesting is mental masturbation ...
Depends on the frequency of trades. Most of our systems are short term; and, therefor can generate a generous sample size quickly,i.e. days, maybe a couple of weeks, never more than a month. Real time testing with real money in small lot sizes is the answer we found. I spent a decade plus designing and backtesting using textbook theories. Never made any real money that way ...
When you're trading a system that was tested like that, at what point do you know you either need to reduce back down to small size or just stop trading it? Do you look at a few key stats and compare the current forward results with the previous forward and backtested stats like: drawdown , win% rate, Avg Profit / Avg Loss, Net Profit? What stats do you look at?
Instead of "they don't work forever" I would say they have periods when they don't work ( nobody knows how long, how deep drawdowns... ). The only reason why mechanical systems work is: Hurst exponent for many markets is above 0.5 which means that the price has a tendency to follow the previous move, rather than reverse. Of course it varies over time, but most of the time is >0.5 Plus you have to use proper position management strategies.
In terms of a purely automated, mechanical approach to a particular trade setup... you're probably right in many cases. In terms of an overall approach to the market... I think that skewing events like the Russian economic meltdown, Brazilian debt crisis, and war/terrorism makes it an extremely risky venture at times. Please refer to Long Term Capital Management. You would think that a mature, highly liquid and fully electronic futures exchange would be fully automated by mechanical systems... but its not the case. As a matter of fact, there's alot of gamesmanship and stop trigger hunting by big locals to take advantage of mechanical trading systems. I use automated order execution systems to implement my strategies at certain times, but I couldn't possibly take a fully automated approach to the market at this point.