Profitability of technical stock trading: Has it moved from daily to intraday data, страница 13

•  Over the last two subperiods the 2580 trading systems (based on 30-minutes-data) would have performed worse than over the entire sample period, between 2004 and 2007 the models would have even produced negative gross returns on average.

The shift in the profitability of technical models from daily data to 30-minutes-data during the 1980s and—hypothetically—from 30minutes-data to data of higher frequencies in recent years could be explained in two different ways.

According to the Adaptive Market Hypothesis (AMH) of Lo (2004), markets become gradually more efficient in an evolutionary process. By learning to exploit profit opportunities, market participants will slowly erode these opportunities through an arbitrage mechanism. Once the “old” and simpler rules have become unprofitable, new and more sophisticated trading strategies will emerge which will gradually also improve market efficiency.

An alternative interpretation is as follows. The continuous rise in the “speed” of transactions in financial markets causes technical traders to use increasingly data of higher frequencies instead of daily data.^[13] As a consequence, intraday asset price movements become more persistent and, hence, exploitable by technical models. At the same time, price changes based on daily data become more erratic which in turn causes daily models to become less profitable. In addition, the use of data of higher frequencies induces traders to use more sophisticated trading models to filter out (very) short-term trends (asset price volatility rises with data frequency). Such a shift will also impact the trending pattern of asset prices (for this feed-back see Schulmeister, 2006, 2007).

The main difference between the AMH and the alternative explanation sketched above is as follows. The AMH assumes that any originally profitable trading rules will become gradually less profitable because more and more people use them. As a consequence, smart traders seek for and finally discover new profitable rules. By contrast, the alternative explanation assumes that the causality runs from the use of new and more complex rules based on an ever increasing data frequency to the erosion of the profitability of the older and simpler rules. This effect is mainly due to the change in the trending pattern of asset prices caused by the gradually increasing use of the new trading strategies.^[14]

To summarize: There are two explanations for why the profitability of technical trading might gradually shift to more complex rules based on data of increasingly high frequencies. The Adaptive Market Hypothesis focuses on the arbitrage mechanism as the driving force of this process, the alternative hypothesis focuses on the self-reinforcing interaction between the type of model as well as the data frequency used, and the specific features of asset price trends. An empirical evaluation of these two hypotheses represents a complex task. Hence, it has to be left to future research.

The resultsof this studydonot imply thattechnical modelsrepresent “money machines” which can easily be run. This is so because technical stock trading—in particular when based on high frequency data— involves different risks which are greater for amateurs as compared to professional traders:

•  Due to the frequent occurrence of “whipsaws,” technical models often produce sequences of mostly unprofitable trades which accumulate to substantial losses. These losses are particularly high if stock futures are traded (leverage effect).

•  Lack of financial resources might also prevent amateur technical traders from sticking to the selected model during “whipsaws” (switching models can easily increase the overall loss).

•  “Model mining” represents a particularly important source of risk. If a technical trader searches for the “optimal” model out of a great variety of trading systems on the basis of their performance in the (most recent) past, then the selected model might suffer substantial losses out of sample if its abnormally high profitability in sample occurred mainly by chance.