Trading systems’ optimization is bad or good ? Let’s recognize the truth, in the trading arena the word “optimization” conveys a negative meaning – the system you are trading is “optimized” so it is untrustful. The system report you are showing around is “optimized” so it is senseless. That software you bought in order to make genetic research is wellworth zero since it just makes “optimization” – you see how many negative meanings the word “optimization” hides inside itself ? If you want to offend a system developer just tell him that he is used to “optimize” his systems and the offence will be direly cruel. But optimization is not as bad as it appears from outside. First of all there is a hefty misunderstanding about optimization since usually it is thought it regards only the input parameters of a system. In reality the parameters optimization is the least devastating form of optimization since it adapts the inputs to the current market activity. If the market is volatile the system should be ready to ride it. It the market is dull the system needs to be cautious about entering too easily a trade. What is really the death kiss in optimization, and nobody seems to be aware of it, is to adapt variables and conditions to a specific market. This is often the case of those systems that are built with thousands of variables and inputs that look more a poem than a programming code. To adapt a trading systems in terms of variables to a market means that every day will have those two conditions that catched just in that occasion the high and low of the day. The following day a new set of conditions will catch the high and the low of the day and so forth. It is a story that statisticians call “degrees of freedom” but you do not need to disturb science to grasp it. Victims of this pitfall are usually academicians, statisticians and over all those people that are professional programmers. They hold the deeply rooted conviction that to develop a winning system you absolutely need to program a complicated, long and serious code. Nothing is so far from reality. Another fatal optimization is to find for every price series and instrument a system among thousands that fits perfectly that price series in that historical period. On the contrary to find a system that performs equally well on different markets and different instruments is quite difficult if not impossible. As an easy example of a system that just work on a single market just take a breakout channel that makes a killing on technological stocks during the 1998 – 2000 price boom. Or a system that favors the short side on stocks during the month of September of the last years included 2001. What can add value to your trading is a system that works equally on the same kind of prices series, that is a system that works fine on almost all stock index futures, another that works on almost all bonds, another that works on almost all currency pairs. Please not that I wrote on “almost” all the markets because each market has its own characteristic due to its time, place, trading venue, players. So do not be worried is a system works on 10 Years Treasury notes and T-Bonds but not on Bund. Let’s say that there should be some other markets where the system makes money beyond the chosen one, but do not expect that it will work perfectly on every market. Trading is not perfection but approximation. And not all prices are the same, there are prices which are more autocorrelated than others, more volatile, more noisy, more dull. In summary if the system is robust and profits are made by a logic driver instead than by chance then what is more dreadful is the optimization by variables and conditions or that one with adaptation to a single market. Inputs optimization is far from being the enemy number one, but traders still need to became aware of it.
Emilio Tomasini is Chief Editor of the Italian edition of TRADERS’ MAGAZINE (www.traders-mag.it), the leading monthly publication in Europe for trading and investment, and since 1999 he organizes the TRADERS’ CUP the most important real money European trading competition (www.traders-cup.it). His personal weekly free newsletter L’Indipendente di Borsa (www.indipendentediborsa.it) is one of the most reputable Italian financial media and counts more than 100.000 readers. His website in English is www.emiliotomasini.com