All systems in every discipline depend on past data. There is no such thing as present data in a system. The instant a present datum is perceived, in the act of perception, it becomes a recorded event, a past fact that has already entered the stream of history.
Therefore, no system, by definition, contains future data. Since the advent of a panic is a surprise, always unexpected by any system, it cannot be anticipated or acted upon before it happens.
But since the panic occurs within time, then no matter what frame of time you design your system to optimize, if you use enough past time frames, your system will include one or more historic panics. That is useful system information.
But practically speaking, you don't want to use old time frames because the evolution of the environment within which you want your system to operate--the present--has changed enough to render valueless or detrimental the inclusion of the old frames in the new system. There is an optimal number of past frames to include beyond which error arrives. (Similarly, there is a minimum number you must use to produce any results which will merit some degree of confidence.)
The time frames on this site range from six weeks to seventeen months. A time frame is the range of time which contains the data that the system processes every time it runs an iteration or an update. The minimum number of iterations to arrive at reasonably reliable results is 200 weeks or more, a testing period of about four years.
THE BASIS OF TRADING SYSTEMS
Whether systems developers know it or not, all trading systems are based on cycles. Cycles are inherent to all phenomena. For the most part, the frequency of the cycles I use are within the time frames I mention above--a few weeks to seventeen months. There is a host of others, both longer and shorter, that I do not use. It is not feasible to attempt to identify all the cycles at play on market prices. Nor is it necessary. A handful dominate. If you get those few right, you will have a profitable system. If it were perfect, you would have no losses or drawdowns. But these occur, and when they do, it is due to the missing cycles you have not included in your system. The options are to 'fix' the system by changing its parts or stay with it, riding through the drawdowns or losses.
Some developers, and their clients, can and do tolerate large drawdowns, believing that the system cycle will turn, and that they will be rewarded by staying with it. That takes varying degrees of patience and fortitude sometimes missing in the psychological makeup of the participants. The other option--'fixing' the system--entails retesting. For example, if you record system results for every 26-week period, say, rolling forward weekly from inception to the present, you will see trends that reflect improvement or disimprovement of results as you modify parts of the system one at a time.
The beauty of market cycles is that you can see with exact precision the instant-to-instant numerical measure (prices) of the human psychology of the crowd who is participating in the market at the time. More beautiful, you can see the dynamic progression of prices from moment to moment as emotions rise and fall. Prices are the interface where emotion meets reason (the normal agent of decision-making to buy or sell). These moving price data are the fuel of the trading system.
OTHER SYSTEMS
Extraordinary. This one combined materials, fuel, propulsion, telemetry, guidance & control, men, women, engineering, math, physics, chemistry, medical science, life support, and procedures. It was designed to land men on the moon and bring them back safely. It succeeded in 1969. Two more extraordinary systems, unique among many: Voyagers 1 and 2, launched as deep space probes in 1977. Since then, each has traveled 12 billion miles toward other star systems and is still going strong after 40 years of continuous daily contact with mission control--including episodes of onboard repairs and alterations directed from earth. The sun is only 93 million miles away. It takes 29 hours each way to 'talk' with Voyager. The complexity of these systems is beyond belief.
Do you not think man is capable of building such a thing as a puny, little stock market trading system? Yes, but it is different from moving hardware through space which draws on the laws of Newtonian physics and Euclidean geometry, nice, tidy tools which precisely measure concrete facts. Trading systems, on ther other hand, require rigorous analysis of numerous, human, free-will decisions which appear to be random. Nevertheless, they leave they leave their own trail of hard, factual evidence--prices. These are the sole basis of a trading system. Their randomness can be tidied into order to reflect the cycles they are expressing. Next, here's an example of someone who has been succeeding in transforming randomness into order and order into profits of stellar magnitudes.
James H. Simons founded Renaissance Technologies Corporation in 1982 to trade investment portfolios solely by mechanical systems . His earnings in
2004 $1.4 billion
2005 $3.1 billion
2006 $3.5 billion
2007 $5.7 billion
Current assets $30 billion.
150 technical staff, half of whom have Ph.D. degress in Math, Physics, Computer Science, Statistics, and Astronomy, continually designs, redesigns, and enhances the system(s). What do you think is in their systems? Cycles, nothing but cycles and their interactivity. Returns average 74% per year (38% per year to investors after managment and incentive fees). $19-million revenue per employee. Fabulous. $10,000 in his Medallion fund in 1989 is now worth $450 million.
In the realm of trading speculation, Jim Simons is the Voyager of deep space travel.
There are two differences between Renaissance Technologies and Copernicus's Winning Systems with ETFs: complexity and frequency. The frequency is subsumed in the complexity. Complexity is inherent in the Universe, as is simplicity. There is nothing wrong with simplicity as long as it contains the minimum of critical parts and processes necessary to fulfill the mission. Frequency is number of trades per unit of time.
Renaissance trades 3000 to 4000 shares of NASDAQ stocks per day to cite statistics on one of their funds. Their holding periods can be as short as seconds. Copernicus trades one to two times per month, holds two positions maxiumum in a portfolio at any one time.
The performance, to the extent that it is available from public sources, and making allowance for differences in portfolio size and length of time in business, has been remarkably close, about 35% to 38% per year at least since 1989 from Renaissance, 37% from Copernicus since 2002. Caveat: Copernicus per year return is based on the average of rolling 12-month returns over six years--Renaissance is not. I suspect theirs would be considerably higher if it were, but do not know for sure.
PROBABILITY AND RESOLUTION
A panic is an in extremis event--pell mell, helter skelter, no resolution possible, or unacceptably delayed.
Since 1790 in the United States, there have been nine weeks out of 11,300 during which past panics started and lasted one month to several years. Probability of occurrence: less than 0.09%. Recovery rate: 100%.
Coda
The Ides of October. Nearby cycles point to the the 14th of October as a day of potential stock market tumult. You know that I eschew forecasting, settling for the more benign, less focussed concept of 'bias' or 'context' (most recently illustrated 