[ Pobierz całość w formacie PDF ]
.4.One cannot know in advancehow hidden software limitations and data adjustments can interfere evenin the case of simple trading systems and alter their logic in an indirect way.Although software programs with back-testing capability give the impres-sion that testing and analyzing trading systems is a relatively easy task, real-ity is different and the whole process can easily turn chaotic.The view thatprofitable trading systems can be developed by just writing a few lines ofcode using a high-level computer language is naïve and experienced tradersknow that is not the case.Learning the proper way of developing prof-itable trading systems while avoiding the pitfalls is a slow process whoseefficiency and effectiveness increase as a function of time and experiencegained from actual trading and by risking real funds under actual marketconditions.In this chapter, the process of development of trading systems via theuse of analysis was discussed and the various steps followed were de-scribed.It was shown that successful analysis relies heavily on ad-hoc waysof discovering trading system models and on subjective analysis of the re-sults obtained from back testing.I have serious doubts that the majorityof trading system developments can be successful with analysis and over-come these serious shortcomings, which are the deeper cause of frequentsystem failures and loss of capital.As an alternative to analysis, I came upwith the process of synthesis of trading systems in the early 1990s, whichis described in more detail in the next chapter.c07 JWBK136-Harris March 20, 2008 9:0 Char Count=CHA PT E R 7Synthesis ofTrading Systemshe most important element of a systematic trading methodology is aprofitable trading system.It is not merely enough to have a high-speedTInternet connection, a platform with direct order execution, fancycharts, real-time news, and analysis tools.Systematic trading requires aconsistent method of generating entry and exit signals that can be imple-mented in a computer language, back-tested, and then, preferably, inte-grated with real-time data and direct execution platforms.Traders whoappreciate the advantages systematic trading has to offer over conven-tional methodologies spend considerable time and effort developing trad-ing systems.This involves, among other things, searching for new ideas,coding the logic of candidate trading systems, and analyzing back-testingresults.All this translates into a time-consuming trial-and-error process.Inorder to test a new idea, a model must be implemented and back testedand the performance results must be analyzed carefully.In many cases,it turns out that the trading system has no chance of producing consis-tent returns in the future because the values of some key performanceparameters are not acceptable.Then, one must start from scratch witha new idea until a system that produces satisfactory performance resultsis obtained.A high percentage of new traders who desire to adopt a systematic ap-proach find the trial-and-error methodology of system development frus-trating.Sooner or later, after a number of unsuccessful attempts to comeup with a profitable system, they give up to adopt other trading styles.Thus,the main cause of frustration is the traditional methodology of systemdevelopment followed, which is fundamentally a trial-and-error process.105c07 JWBK136-Harris March 20, 2008 9:0 Char Count=106 PROFITABILITY AND SYSTEMATIC TRADINGIn addition, this methodology requires a source of new ideas in order tokeep the trial-and-error process going with the hope of a final convergenceto an acceptable system.But even in the presence of a source of new ideas,besides being time consuming, the process of testing and analysis has sev-eral pitfalls, as already discussed in Chapter 6.The realities of traditional modeling, testing, and analysis used for thepurpose of developing systematic trading systems can be overcome anddealt with only by a small percentage of traders who possess the skills andexpertise required for accomplishing their objectives.The rest are discour-aged from the start because they lack the skills and experience.Partingwith a few thousand dollars to purchase user-friendly software packagesfor trading system development does not make things any easier.Knowl-edge of software programming is still a prerequisite for developing sys-tems, even when using those user-friendly programs.Abstract ideas mustbe translated into a mathematical model and then into some custom high-level computer code.Thus, the requirements point to an unrealistic taskfor the majority of traders.FROM ANALYSIS TO SYNTHESISOne way of dealing with the harsh realities of the traditional process ofdeveloping trading systems is the concept of synthesis.The term synthesisis defined as a systematic process by means of which trading systems thatfulfill user-defined performance criteria and risk/reward parameters can bediscovered in a fully automated fashion.In order to understand how such a concept might work in principle,let us first review the traditional trading system development methodol-ogy shown in Figure 7.1.According to this widely followed methodology,a model of a trading system must be identified in advance by some means.The identification process can be based on empirical rules, chart analysis,technical analysis indicators, traditional chart pattern formations, candle-stick pattern formations, and so forth.After the model is identified, it must be coded in a computer languageso that its historical performance can be tested.This is known as theimplementation step.Many developers use a high-level programming lan-guage for this step, which is available as part of a commercial softwareprogram specifically designed for this purpose, but others prefer to writecustom code.The back-testing step involves driving the model with historical datainput to affect generation of market entry and exit signals followed by thecalculation of a set of performance parameters based on the price levels atc07 JWBK136-Harris March 20, 2008 9:0 Char Count=Synthesis of Trading Systems 107HistoricalModel ModelBack TestingIdentification ImplementationDataEntries PerformanceExits PerametersModelAnalysis of Results Back-Test ValidationImprovementNOAccept?YESSave and UseFIGURE 7.1 Traditional methodology of trading system development.which the signals occurred.A partial list of some important performanceparameters was given in Chapter 6, Table 6.1.As soon as the back-test results are obtained, a manual validation of-ten follows in order to verify that the trading system operates according tospecifications.If the results are successfully validated, then a performanceanalysis is carried out to determine whether the trading system should beused in actual trading [ Pobierz caÅ‚ość w formacie PDF ]