Trading Systems

A new approach to system development and portfolio optimisation

Emilio Tomasini, Urban Jaekle

• Publisher: Harriman House
• Year: 2011
• ISBN:
  9780857191496
• Topic: Personal Development
• Subtopic: Personal Finance

"The key is how to adapt existing codes to the current market conditions, how to build a portfolio and how to know when the moment has come to stop one system and start another one."Every day there are traders who make a fortune. It may seem that it seldom happens, but it does - as William Eckhardt, Ed Seykota, Jim Simons, and many others remind us. You can join them by using systems to manage your trading.This book explains exactly how you can build a winning trading system. It is an insight into what a trader should know and do in order to achieve success in the markets, and it will show you why you don't need to be a rocket scientist to build a winning trading system.There are three main parts to Trading Systems. Part One is a short, practical guide to trading systems' development and evaluation. It condenses the authors' years of experience into a number of practical tips. It also forms the theoretical basis for Part Two, in which readers will find a step-by-step development process for building a trading system, covering everything from initial code writing to walk forward analysis and money management. Part Three shows you how to combine a number of trading systems, for all the different markets, into an effective portfolio of systems.A trader can never really say he was successful, but only that he survived to trade another day; the "black swan" is always just around the corner. Trading Systems will help you find your way through the uncharted waters of systematic trading and show you what it takes to be among those that survive. A new approach to system development.

Part I. A Practical Guide to Trading System Development and Evaluation

1. What is a trading system?

Nowadays the term trading system conveys many meanings that can sometimes be misleading. A trading system is a precise set of rules that automatically defines, without any human discretionary intervention, the entry and the exit on the markets. Since rules are precise there is no doubt over when and where to apply them and this makes the trading system statistically testable. This means that we can figure out how the system performed in the past and how it could perform in the future with a certain degree of confidence. If you add a money management rule and a portfolio rule to the set of rules that define entry and exit on the market then you have a “trading strategy” or, in other terms, a completely automatic approach to the markets, given a starting capital.

When we talk about money management we are not talking about what is commonly believed to be risk management; that is, where to place an initial stop loss or a target price and so on. We are talking instead about “how much” to invest on a particular trade; that is, the position sizing or how many shares and how many futures contracts to buy and sell. And if we move to the construction of a portfolio of systems on uncorrelated price series, then money management is foremost what we should deal with in order to maximise the portfolio returns relative to the risk. Thus this process is also called portfolio management.

In more practical terms we can conclude that in order to develop and implement a trading system you need to have a software that easily performs all the programming and testing facilities and above all that goes directly to the market without any interference by the user. So we need to distinguish from a purely linguistic standpoint what a trading system is (or algorithmic trading) and what automated trading is. Indeed the latter could not exist without the first, but not vice versa. You could have algorithmic trading signals provided by a computer but not automatically place them on the markets. The main hindrance for the trading systems user is to produce trading signals that he is not mentally fit to place in the marketplace. To have a trading platform which automatically trades the mechanical signals produced by the trading system is thus a major advance.

1.1 An easy example of a trading system

So what does a trading system’s pseudo code look like? It could be the following:

Buy 2 contracts at the highest high in the last 20 days;

Sell short 2 contracts at the lowest low in the last 20 days;

If marketposition = 1 then sell at last close – avgtruerange(14) stop;

If marketposition = -1 then buy to cover at last close + avgtruerange(14) stop.

We have an entry rule and we have a stop loss rule. This is a trading system. Its risk management is quite poor since we just wrote an initial stop loss which works also as a trailing stop, but the example is easy and quickly shows what a trading system is.

When the investor or the trader has a predefined set of rules that she or he applies discretionally in order to enter or exit the market, without any testing process and without any automation of the orders, and resting on a final judgment if and when to enter or exit the markets that could not be eventually classified ex ante, we could more appropriately talk of a “trading methodology”. If the investor or the trader conducted detailed research on the past behaviour of the trading methodology, supporting it with statistical tests, and he has a disciplined character so that all the signals are equally placed on the market, we have something that is much closer to a trading system, without, in any case, being it. Since the “trading system” is much more precisely identified and it conveys an idea of a scientific work that underwent a strict statistical test, many investors or traders are tending to profess the use of a trading system instead of a trading methodology. A trading methodology always involves a bit of judgment and discretion.

Recently the financial industry has been swamped by the “quants”, that is by money managers and traders that apply quantitative methods in order to produce buy and sell signals. What the difference is between a trading system and a quantitative forecasting method nobody knows, but since the term “quantitative finance” conveys an idea of something which is rigorously scientific and surely beyond the retail-oriented trumpery wares of common technical analysis, expect to meet many system traders that resell themselves as “quantitative traders”. If the term “quantitative finance” serves to divide the system traders that base their decisions on statistics from those analysts that just grasp the artistic and esoteric side of technical analysis, we all agree on calling ourselves “quantitative traders”.

Since a scientific appeal is the best way to sell something, there is nowadays a wide rush in the markets to give a deep scientific status to the trading systems industry. This approach seems to take for granted that a trading system must be a long series of rules, programmed in a complicated way, and full of breathtaking algorithms. Salesmen know very well that complexity raises prices. But it also raises the probability that a trading system will fail in the real world, and there is no approach more false than this.

Many commercially available formulas you can find in any technical analysis software, when properly tested and applied to price series, show a real market bias; that is they have a trustful predictive power. Trading systems could be very easy in their logical implementation, like a channel breakout, an indicator, a moving average, and to rely on your own trading decisions on something that is “easy” will not reduce your success probability; on the contrary it will increase it. On the way to success a lot will be done by money management and portfolio construction, risk management and timeframe, so do not be worried when you examine a trading formula that is simple and produces an equity line that appears unexciting, because you need to always reason under the portfolio constraint. It must be clear from the very beginning that a mediocre trading system – if applied to a portfolio of markets – will easily produce a good looking equity line.

1.2 Why you need a trading system

A huge economic literature shows without any doubt that just a few percent of traders are able to beat the market year after year. Most of both the retail and institutional traders sooner or later will go bust. If you do not belong to the lucky category of winning discretionary traders, then the only option for you in order to survive is the use of trading systems. If you have purchased this book you are most likely not a successful discretionary trader: in my experience successful discretionary traders are intuitively blessed and are unconsciously able to predict the market moves with their gut feeling. On the contrary there are many successful money managers, institutional and retail traders that profit from predetermined trading strategies and investing methodologies. But it would be misleading to think that a trading system could easily overcome all the hindrances trading creates. A trading system from one side could help the trader to beat the market but from the other side will create a new set of problems that a discretionary trader does not know.

First of all, if the trader has problems in terms of physical courage and some difficulties in pulling the trigger, trading systems will not be the ultimate solution. Like Larry Williams says, ‘trading systems work, system traders do not’. There is no bigger infamy for a systematic trader than not to take a signal, as Bill Eckhardt wrote:

If you make a bad trade, you have money management, you have a whole bunch of things that will come to your aid, and you’re really not in so much trouble if you make a bad trade. But if you miss a good trade there’s really nowhere to turn. If you miss good trades with any regularity you’re finished, you’re doomed in this game.

Second, in order to trust a trading system, especially during gloom periods when drawdown will erase the trader’s confidence in the trading system’s capabilities, you really need to do a huge amount of research and statistical work that not everybody is able to do. To develop, implement, test and evaluate a trading system is not an everyday job.

Finally, many of the drawbacks that affect discretionary trading still affect systematic trading, e.g. lack of sufficient starting capital, possibility to diversify the portfolio, full-time, 24-hour, dedication.

More importantly we can say that trading is not a rational enterprise, it is not an activity where you can, given some premises, arrive at a unique conclusion or where everything could be explained in a logical way. Fear and greed manipulate prices in a way that the human mind is unable to grasp. There are of course some fortunate discretionary traders that can beat the market with gut feeling but in these instances they do not often manage to fully explain why they buy or sell. If all this is true the consequence would be that you need a tool that is not rational and not logical to enter and exit the market, something that you could not fully understand, something that is counterintuitive. Usually the signals that you believe to be illogical or simply prone to failure will be the big winners.

To use a mechanical trading system means that you need to discard widely held beliefs about finance and over all discard the “feel-good” approach to trading: everybody usually feels comfortable buying dips and uncomfortable buying the highest high, but it may be the case that just the latter methodology is the good one. Testing a trading system could mean being forced by the brutal power of numbers to a trading attitude where you do not feel at ease. To be a fully mechanical trader means, in conclusion, to use violence against yourself. This is the only way to profits, unless you are one of the fortunate gun slingers that make money day after day and do not even know how.

1.3 The science of trading systems

It would be inappropriate to mix all the kinds and breeds of technical analysis available nowadays. There is a broad distinction between subjective and objective technical analysis.

Objective technical analysis methods are well-defined repeatable procedures that issue unambiguous signals. This allows them to be implemented as computerised algorithms and back tested on historical data. Results produced by a back-test can be evaluated in a rigorous quantitative manner. Subjective technical analysis methods are not well-defined analysis procedures. Because of their vagueness, an analyst’s private interpretations are required. This thwarts computerisation, back testing, and objective performance evaluation. In other words, it is impossible to either confirm or deny a subjective method’s efficacy. For this reason they are insulated from evidentiary challenge.[2]

Subjective technical analysis did not gain a good reputation among the academic community or among serious market practitioners because of its vagueness and lack of scientific method. To be a chartist or a technical analyst, instead of a portfolio manager, using hidden technical analysis could be the least promising career launching pad in the financial industry. There is a wide sociological and psychological literature about why people believe weird things such as dogma, faith, myth and anecdotes so that it is much easier to isolate good scientific technical analysis using the approach of the scientific method. Without the intention to lecture here about philosophy of science we need to briefly remind readers what scientific knowledge is. Scientific knowledge is empirical or it is based on observations of reality:

‘The essence of technical analysis is statistical inference. It attempts to discover generalisations from historical data in the form of patterns, rules and so forth and then extrapolate them to the future.’[2]

So technical analysis, utilising the tools of statistical inference, starts from a sample of observations in order to gauge some statistical properties of the whole population. In this way technical analysis, like statistics, is quantitative. Further, technical analysis, like science, tries to predict the future through functional relationships among variables. If this variable does this then the dependent variable will do that. What in technical analysis is a rule in statistics is a functional relationship, which a certain probability is attached to. There is no barrier between scientific technical analysis and statistics so that all the doubts raised by those that dislike subjective technical analysis suddenly disappear. Quantitative technical analysis uses the typical method of analysis in applied sciences: the hypothetic-deductive method initiated by Newton and made famous by Popper. The hypothetic-deductive method has five stages [2]:

1. Observation. The system developer, through the continuous observation of the daily and intraday activity of the financial markets, devises a relationship among variables, i.e. among the daily volume activity and the closing price, or among the value of an indicator and the next day opening.
2. Hypothesis. This comes from the innovative mind of the system developer – an intellectual spark, the origins of which nobody knows. The system developer understands that the relationship he hypothesises is not due by chance to the particular nature of the sample he analysed, but it is common to the majority of the samples he can deduct from the whole population of data.
3. Prediction. If the relationship is true then a conditional proposition or a prediction can be constructed and ‘the prediction tells us what should be observed in a new set of observations if the hypothesis is indeed true’. ^[2]
4. Verification. The system developer verifies if the prediction holds true in a new set of observations.
5. Conclusion. The system developer, through the use of statistical inference tools such as confidence intervals and hypothesis tests, will decide if the hypothesis is true or false weighing whether new observations will confirm the predictions.

This process is in no way different from the scientific appraisal method used in applied sciences like chemistry or biology.

Endnote

2 Aronson, David, Evidence-Based Technical Analysis, Wiley, 2006, see [2]. [return to te...