In this chapter we give a brief introduction to different methods that may help to improve the performance of your portfolio: technical analysis, neural networks and log-optimal portfolios. The common idea behind these methods is that past price movements may help in forecasting future trends. In other words, we implicitly assume that prices do not follow a Markov process (for example random walk), but they have some kind of long lasting memory, hence patterns from the past may reoccur also in the future, all in all markets are not efficient.
In the first part we introduce the most common tools of technical analysis and present some indicative examples of how to program them in the R environment. In the second part we outline the concept of neural networks and their design by R's built-in function. Technical analysis and neural network are applied on the bitcoin database, thus we focus on a single asset and investigate for reliable signals of buying and selling. Finally, in the third part we discuss the so called log-optimal portfolio strategies that enable us to optimize our portfolio of several assets (in our example some NYSE stocks) for the long run.
The main goal of this chapter is just to give a helicopter view on the concepts, the most common tools that are used and to provide some examples of their programming. Therefore we would like to underline here that, by need of being concise, we only intend to give you some insight into the field and to entice you to check the references, learn more and try further tools yourself.
Markets are efficient to the extent that all information is built into the current prices. The weak form of market efficiency requires that the latest price already incorporates all the information which can be obtained from the chart of past prices and trading volumes. Clearly, if markets were efficient at least in this weak sense, returns would be totally independent over time and strategies based on technical analysis, neural networks and the logoptimal portfolio theory would be completely worthless, see Hull (2009), Model of the behavior of stock prices.
However, the efficiency of a given market is purely an empirical question. You can never be sure that asset returns in the real world are really completely independent in time. Therefore, you should not take market efficiency as a fact but you are encouraged to test it on your own by inventing and implementing new technically inspired strategies. If your strategy calibrated on past trading data proves to be robust enough and works well in the future, then the market will generously honor your efforts by enhancing the risk/return profile of your portfolio, and, as a result, you will earn an extra profit. Studies have shown that emerging currency markets, for instance, are less efficient due to illiquidity and to central bank interventions, see Tajaddini-Crack (2012); whereas most technicist strategies do not hold on the more developed American stock market Bajgrowicz-Scaillet (2012), Zapranis-Prodromos (2012). Furthermore, the same studies indicate that when technical trading is successful, its combination with fundamental analysis is even more so. Zwart et al. (2009).
Despite being sort of an apocrypha still today, technical analysis is widely used even among fundamental investors. This is mainly due to its self-fulfilling nature: as market players know that more and more of their peers are using the TA tools they keep an eye on them, too. If, for instance, a 200-day moving average is breached on a main index chart, it is likely to make the headlines and cause a selling wave.