How Do I Design a System?

Now you are convinced that you need to design a system for trading. However, how do you do that? Let us look at some of the requirements and steps involved in creating an effective system.

1. Requirements for Designing a System

What is needed to design a successful system? Before even considering the components of a system, we must begin with something even more basic—designing a workable, profitable system begins with some basic personal attitudes. Some of the characteristics of the necessary mind-set include the following:

• Understand what a discretionary or nondiscretionary system will do—be realistically knowledgeable, and lean toward a nondiscretionary, mechanical system that can be quantified precisely and for which rules are explicit and constant.
• Do not have an opinion of the market. Profits are made from reacting to the market, not by anticipating it. Without a known structure, the markets cannot be predicted. A mechanical system will react, not predict.
• Realize that losses will occur—keep them small and infrequent.
• Realize that profits will not necessarily occur constantly or consistently.
• Realize that your emotions will tug at your mind and encourage changing or fiddling with the system. Such emotions must be controlled.
• Be organized—winging it will not work.
• Develop a plan consistent with one’s time available and investment horizon—daily, weekly, monthly, and yearly.
• Test, test, and test again, without curve-fitting. Most systems fail because they have not been tested or have been overfitted.
• Follow the final tested plan without exception—discipline, discipline, discipline. No one is smarter than the computer, regardless of how painful losses may be and how wide spreads between price and stops may affect one’s staying power.

2. Initial Decisions

Once you are committed to the mind-set and discipline of creating a system, you must make certain decisions about the characteristics of your system. The actual fundamental or technical method used as the basis for the system is relatively unimportant. What is important is that whatever is used can be defined precisely. Most fundamental and technical methods, by themselves, have a sketchy record of performance. Performance in the system will depend more on filters, adjustments, and the entry and exit strategies than the method itself. This does not mean that any old method will work. Pick a model (entry and exit method that has some statistical probability of success) that is familiar, sensible, comfortable, and has a decent record. Be sure it is based on facts, not opinion, and then concentrate on the process of developing a system.

Most systems designers argue that the simpler the system, the better. A system can become bogged down with large numbers of conditions and statistically will lose degrees of freedom, requiring more data and more signals to establish its significance. The market has entropy, an inherent disorder that changes periodically in unexpected ways. A system with few variables will reflect the patterns in the market with a certain accuracy. As more variables are added to the system, entropy causes the nonpattern variables to increasingly influence the results, causing the system to eventually decrease profitability because it can account only for the patterns but not the internal market changes. Indeed, when testing a system, the added variables should be tested for their effect on the system results, and if the performance declines, those variables should be eliminated even if they appear logical. Some designers such as Richard Dennis argue against simplicity (Collins, 2005), but they have enormous computer power and knowledge behind them. Hill and others argue that even with modern technology and mathematics, the success of systems now is no greater than the classic systems designed with a hand or crank calculator.

First, you must decide what kind of input and tested model is to be used to generate signals. Some investors depend on fundamental information; most traders depend on technical methods. Others use a combination. The important aspect is to have a clear understanding of the system’s premises and to know that the rules will be easily quantifiable and precise. Specificity is much easier to use and to test than generality. You must also understand the logic of the system and be sure that it suits your style of trading or investing.

Second, you must decide on which markets to focus. Is the market suitable for the intended system? Are there opportunities for diversification between markets or instruments? How much volatility and liquidity is required, and what specific instruments will be traded?

Third, you must establish the time horizon for the system. For example, most trend-following systems work better over longer periods, but most pattern systems work in hours and days. Does the system intend to scalp trades, swing trade, or long-term invest? In addition, what is the psychologically best-suited time not only for system logic but also for ease of use? Do you have time to spend all day with the system, or can you monitor the system only daily, weekly, or monthly?

Fourth, you must have a risk control plan; otherwise, you will not know what to do when markets change. Understand that losses are inevitable, but be sure to keep them under control. Admitting losses separates the professional from the amateur. Rationalizing or excusing losses never helps. The market is never wrong—get out, the quicker the better. To do this, devise a stop-loss strategy—“no clinging to the mast of the sinking ship.” This strategy should include protective and trailing stops, price targets, and adjustments for volatility, type of market, and any other state that the market might be in. Another option is to have a filter that shuts down the system when the market enters a trading range or has other characteristics that detract from the model’s performance. Otherwise, the account may suffer a larger loss. Emotions and judgment become adversely affected, causing missed opportunities, selling profitable positions to get even, and other mistakes. Stop-losses free up nonproductive capital and cause less stress once accepted. In addition to risk control, you must decide whether you should use leverage or pyramiding.

Fifth, establish a time routine, which should include when to update the system and necessary charts, plan new trades, and update exit points for existing trades. As part of your system administration, maintain a trader’s notebook, a trader’s diary, and a daily equity chart. Maintain a daily trading sheet (similar to an accounting ledger) and a position sheet.

3. Types of Technical Systems

Technical analysts use a number of types of technical trading systems. Although there are numerous systems, they can be divided into four main categories: trend following, pattern recognition, range trading, and exogenous signals systems.

3.1. Trend Following

From our knowledge of technical systems, we understand that markets trend at times and trade in a range at other times. The most profitable background is a trending background because the moves are larger and generate fewer transaction costs. While periodically trend trading becomes difficult and many traders begin to believe it is dead, it is not. As Bill Eckhardt, partner of Richard Dennis and originator of the Turtles, has been quoted, “I have lived through the death of trend-following a half dozen times, and, like Mark Twain’s death, it was highly exaggerated.” (Collins, 2005). Most large-scale mechanical system hedge funds and commodity trading advisors use trend-following systems. Rather than attempting to catch the peaks and valleys, the trend­following system acts in the direction of the trend as soon after it has begun as can be reliably detected. Contrary to the buy low and sell high philosophy, the trend-following system will buy high and sell higher. Schwager believes that slower, longer trend systems work better because the gains are larger, although less frequent, and the whipsaws are minimal. Most trend-following systems add a trend indicator such as the ADX to their set of rules to be sure that a trend is in existence. As we know from earlier studies on trends, the performance of a trend-following system can suffer during a trading range market.

3.2. Moving Average Systems

The classic trend-following system is composed of two moving averages that generate signals when they cross over each other. In his book The Definitive Guide to Futures Trading, Larry Williams discusses how, as early as the 1940s, Donchian demonstrated the validity of this method and showed that it was more successful than the older system of using price versus a single moving average.

If two moving averages are better than one, would three be even better? No, studies have shown that adding more moving averages weakens performance because of the increased number of rules required. Although practitioners frequently report success using moving averages, we must mention that academic studies have shown that moving average crossover systems, even with simple filters, are generally unprofitable. However, academics have not used any kind of risk control in their experiments. Without the use of these important risk-control strategies, the academic studies are not a true measure of the profitability of using a moving average crossover system.

3.3. Breakout Systems

A variation of the trend-following system is the breakout system. These systems generate buy and sell signals when price moves out of a channel or band. The most popular of these systems is based on a variation of the Donchian channel breakout system or some kind of volatility breakout system using Bollinger Bands or other measures of range volatility. The breakout system can be long term and use weekly figures, or short term, such as the open range breakout systems used intraday.

3.4. Problems with Trend-Following Systems

Given their profitability, the moving average and breakout strategies are popular. Because many of these trend-following systems are being traded, many others will receive the same signal at roughly the same time and price you will. Liquidity can become strained, and slippage costs from wider spreads and incomplete fills will increase the transaction costs over what may have been anticipated. The solution to this problem is to devise an original system or to spread out or scale entry orders.

Another problem with trend-following systems is that whipsaws are common, especially during a trading range market, as the system attempts to identify the trend. In fact, trend-following systems often produce less than 50% wins because of the many whipsaws during ranging markets. This problem can be reduced with the use of confirmations, such as special price requirements (penetration requirement, time delay, and so on), once a signal has been given, or through filters and diversification into uncorrelated markets.

Inevitably, to avoid whipsaws, a trend-following system will be late in the trend and will thus miss profit potential at both ends of the trend. Unfortunately, this is the cost of a trend-following system. If an attempt is made to clip more profit at each end of the trend, the number of losses will increase from the ranging nature of the trend at its terminal points. On the exit side of a trend, specific trailing stops or such can be used to receive better prices, but again there is the risk of missing another leg in the trend by exiting prematurely.

Losses occur primarily in the trading range preceding the establishment of a trend, as the system tries to identify the next trend as closely as possible. One strategy to combat this is to use a countertrend system at the same time, even if it is not as profitable as the trend-following system. The gains from the countertrend system will offset some of the losses of the trend-following system, and the overall performance results will improve over the trend-following system alone.

Moving-average and breakout systems are usually limited to a one-directional signal only. Part of the advantage in following a trend is to pyramid in the direction of the trend as evidence of its viability becomes stronger. To accomplish this in a trend-following system, other indicators must be used, thus increasing the complexity and decreasing the adaptability of the system.

The greatest fault with trend-following systems is the large percentage of consecutive small losses that produce significant drawdowns. For example, let us say that the system suffers ten small losses in a row while in a trading range. The drawdown to the equity of the account accumulates during this period from the peak of the equity to the subsequent cumulative loss. A series of losses that cause a large drawdown affect not only the pocketbook but also the confidence in the system and often lead to further complications. One strategy to lessen a sequence of losses is the strategy mentioned previously of using a countertrend system. Another is to initiate only small positions on a signal until the trend is well established. Yet another is to run another trend­following system parallel that has a longer or shorter period.

Because a trend-following system often is characterized by clumps of large profits from the trend and many small losses from the trading range, extreme volatility occurs in equity. We will look at this later when we study equity curve smoothness, but the most-often-used countermeasure is to diversify into other markets or systems.

As with most mechanical systems, a trend-following system can work well during testing and then bomb in practice. In most cases, this is due to improper testing and adjusting. Sometimes the improper testing is due to unrealistic assumption about transactions costs. Unrealistic assumptions including spreads during fast markets, limit days in the futures market, and other possible anomalies may have given false results during the testing stage of the system under consideration. Remember that the popularity of trend-following systems can affect slippage; this fact often is erroneously ignored in the testing phase.

Occasionally, substantial parameter shifts will occur that the adaptive system will not be able to recognize and accommodate. Again, by diversifying by using more than one system or using market character adjustments to volatility, such problems can be reduced.

4. Pattern Recognition Systems

“Every ship at the bottom of the sea had plenty of charts” is attributed to noted systems trader Jon Najarian (Patel, 1997). Using patterns requires considerable testing and overcoming the problem of defining patterns. Larger patterns do not succumb to easy computer recognition because of their variable nature. System traders such as Larry Williams, Larry Connor, and Linda Raschke use short-term patterns, some of which we discussed in Chapter 17, “Short-Term Patterns,” and limit their exposure with specific position stops and price or time targets. Generally, such systems are partially discretionary because they require some interpretation during the trade entry.

5. Reversion to the Mean

Reversion to the mean systems are based on the buy-low-sell-high philosophy within a trading range and are also called trading range systems. This type of system requires a certain amount of volatility between the peaks and valleys of ranges; otherwise, transaction costs, missing limits, and being stopped out on false moves chew up any potential profits. Generally, these systems are discretionary. They profit from fading small counter-trend moves or moves within a flat trend and using oscillators such as the stochastic, relative strength index (RSI), the Moving-Average Convergence/Divergence (MACD), or cycles. The largest potential problem in trading with one of these systems is the possibility of a trend developing that creates the risk of unlimited losses. Protective stops are a necessity.

Generally, this type of system does not perform well. A number of publicly available tests—for example, of buying and selling within Bollinger Bands—have been conducted, and invariably the best performance comes from buying and selling on breakouts from the bands rather than trading within them. The major use of countertrend systems is to run coincident with trend-following systems to dampen the series of losses in the trend-following system during a trading range.

6. Exogenous Signal Systems

Some systems generate signals from outside the market being traded. Intermarket systems, such as gold prices for the bond market, would be an example of an exogenous signal system. Other examples are sentiment such as the VIX for S&P futures, volume, or open interest warnings of activity that trigger price systems or act as confirmation of price systems, or fundamental signals such as monetary policy or consumer prices.

7. Which System Is Best?

Which type of system is the best? John R. Hill and George Pruitt, whose business is to test all manner of trading systems (www.futurestruth.com), maintain that the best and most reliable systems are trend-following systems. Within trend-following systems, the breakout systems have the best characteristics—specifically the Bollinger Band breakout systems, and the Donchian, or channel, breakout systems. Closely behind are the moving-average crossover systems.

Source: Kirkpatrick II Charles D., Dahlquist Julie R. (2015), Technical Analysis: The Complete Resource for Financial Market Technicians, FT Press; 3rd edition.