A few times I do really hope that I know what I'm talking about, but this time I don't. Just kidding!
You'll find each of these types of trades everywhere at some time or other to some degree or other. However, breakouts and retracements are more likely to be successful in trending markets that have enough range to make the trade worthwhile. In a choppy market, reversals have a higher probability of success, though unless one isn't doing any scaling at all, at least one contract can be held in the event of a trending move.
I don't think that's strictly true. There are techniques more sophisticated than merely measuring momentum and trading with or against it.
this topic interests me because i have been trading one fairly simple mechanical system for most of this year and done well with it. my concern is the uncertainty over whether it could either make a small fortune over a period of years, or peter out for good, starting tomorrow. i have several decent backups and variants to fall back on, but they are a bit too closely related for comfort. unfortunately, i realize i can't solicit specific advice and opinions without disclosing details ... but i would be interested in anyone's general thoughts/experiences on the maturation of mechanical systems, or war stories regarding great systems that died. the turtle system is an excellent example, thanks runningbear. without inviting the system sellers and uber-gratification-seekers, has anyone traded a great simple system that did last several years and compound as predicted?
Bob, Thats funny. I have something vary similar but it holds for two days. Has a 70% chance of being correct. What I do to give myself a good fealing is to test random groupings of stocks for each year. For instance ill take say 50 stocks in 95-96 and see how it did. If the system has held up for the past 10 year - 20 years what makes you think its just going to stop today?
Most of my systems are designed using 10 years of back data, and I expect them to hold up for several years going forward without any tweaking. One of my daytrading systems that I use now is essentially the same as the model I was using 10 years ago, apart from a few refinements here and there over time. I've also got a real outperforming system that is only tested back 18 months. It has a sizeable edge and I don't know how long it will last, so I am trading it at maximum size to get the most out of it while it still works. There is no real way of knowing how long your system will hold up. What I do is have a portfolio of uncorrelated systems, and only trade each model while it is making money. I only need 20-30% of my systems to perform as well as historical simulations to make an acceptable portfolio return this way.
Really good system will last forever. It has to have built in features that will adjust to recent conditions and characteristics of price action . It require more work to generate signals then MA crossover system, but it will last forever. Walter
I can tell you of one system which will work for all of you most of the time, in fact I should market it!!!!! All you have to do is observe me, When I buy that is a guarantee that the stock will plummet and you should therefore go short. Observe me when I sell because that is an indicator that the stock is about to take off and you should buy. Tongue in cheek but if you had followed these rules over the last 3 weeks you would have made a lot of money !!!!
In theory yes, in reality no. While self-adjusting parameters will allow a system to compensate for normal variations of the underlying, if a fundamental change in the underlying occurs, it can likely effect a system beyond help. An example is volatility or liquidity. If either shrink to a point where the system's edge will not overcome slippage/transactional costs, the system is doomed. Another example is introduction of new technology. When Globex was introduced to the S&P futures, the fundamental trading characteristics of the pit contract were changed forever. Many trading methods simply would not work anymore while other methodologies were spawned. Monitoring of a systems real-time or out-of-sample results vs. it's historical performance is critical. Large variations in drawdown and/or profit of RT vs. historical are good indications a system might be failing. Best and cheers. rcreal