آیا معاملات با فرکانس بالا تجزیه و تحلیل فنی و کارایی بازار را تحت تاثیر قرار می دهد؟ و اگر چنین است، چگونه؟

In this paper we investigate how high frequency trading affects technical analysis and market efficiency in the foreign exchange (FX) market by using a special adaptive form of the Strongly Typed Genetic Programming (STGP)-based learning algorithm. We use this approach for real one-minute high frequency data of the most traded currency pairs worldwide: EUR/USD, USD/JPY, GBP/USD, AUD/USD, USD/CHF, and USD/CAD. The STGP performance is compared with that of parametric and non-parametric models and validated by two formal empirical tests. We perform in-sample and out-of-sample comparisons between all models on the basis of forecast performance and investment return. Furthermore, our paper shows the relative strength of these models with respect to the actual trading profit generated by their forecasts. Empirical experiments suggest that the STGP forecasting technique significantly outperforms the traditional econometric models. We find evidence that the excess returns are both statistically and economically significant, even when appropriate transaction costs are taken into account. We also find evidence that HFT has a beneficial role in the price discovery process.

The extensive use of technical trading rules by currency market practitioners has long been a puzzle for academics. On the one hand, as Cheung and Chinn (2001) and Gehrig and Menkhoff (2003) note up to 40 per cent of foreign exchange (FX) traders worldwide rely on technical analysis as their main trading tool. On the other hand the Efficient Market Hypothesis (Fama, 1970) suggests that in a market with vast trading volume and virtually non-existent private information about fundamentals, such as the foreign exchange (FX) market (turnover of 2000 billion US dollars per day; BIS, 2005), trading rules based on historical price information should not yield excess profits to traders. Most academic studies related to technical trading in the FX market are inconsistent with real-life practice because they largely limit their trading strategies to daily data observations (Brabazon and O’Neil, 2004, Qi and Wu, 2002 and Reitz and Taylor, 2006). However, nearly all FX traders who use technical analysis operate at a high frequency (Gomber et al., 2011, Ahlstedt and Villysson, 2012 and Guo, 2012). In addition more than 75 per cent of FX trading has been shown to take place within a single day (BIS, 1996), and that the applicability of technical analysis increases with the frequency of trading (Taylor and Allen, 1992). While some empirical studies of daily FX data report the existence of significant profits (Martin, 2001, Mathur et al., 2001 and Saacke, 2002), some other studies demonstrate the contrary (Levich and Thomas, 1993, Lee and Mathur, 1996 and Lee et al., 2001). Studies on the profitability of intra-daily technical analysis also do not convey a clear picture. Some authors report significant net profits (Gencay et al., 2002 and Gencay et al., 2003), whereas others find technical trading to be unprofitable even at these high frequencies (Cucio et al., 1997, Osler, 2000 and Neely and Weller, 2003). Moreover, studies on FX technical trading profitability typically fail to account for transaction costs, trading rule optimisation over time, out-of-sample verification, and data snooping issues (Park and Irwin, 2007). So far, our discussion has focused on the relationship between high frequency trading (henceforth HFT) and technical analysis. However, the question could be reversed and the impact of HFT on the market's quality can also be investigated. Some empirical and theoretical studies suggest that HFT improves market liquidity, reduces trading costs in the form of narrower bid-ask spreads, and makes stock prices more efficient (Jones, 2013). On the other hand, the empirical evidence is somewhat mixed and there are theoretical arguments that HFT can have negative effects. The speed of trading could put slower moving market participants at a disadvantage, leading to adverse selection and reduced market quality. Chordia et al. (2013) argue that buy-side investors could struggle to trade large positions, and their speed disadvantage reduces their ability to supply liquidity leading to increased costs. Chaboud et al. (2010) provides evidence that computer trades are more highly correlated with each other than human trades, indicating that strategies generated by machines are not as diverse as those developed by humans. There is also a possibility of an unproductive arms race developing with HFT institutions competing to be fastest (Jones, 2013). The substantial investments in computer and communication power necessary to reduce latency in trading poses the question of whether HFT adds value overall (Chordia et al., 2013). Given the lack of conclusive evidence on its impact policymakers around the world are still debating whether to introduce limits on HFT or even to completely ban it. Academic studies to date have mainly analysed stocks particularly the 120-stock NASDAQ HFT dataset. We contribute to this debate by examining the FX market where the ability to observe all trading in our experiments allows us to investigate the impact of HFT bid and ask orders on market quality. In this study, we implement a special adaptive form of Genetic Programming (GP), called Strongly Typed Genetic Programming (STGP). The advantage of STGP over the conventional Genetic Programming (GP) used in most previous studies is that STGP evaluates the fitness of agents through a dynamic fitness function which processes the most recent quotes of the six currency pairs in our experiment, rather than a re-execution of the same trading rules. Despite the voluminous literature on the topic, no other study has combined implementing the STGP technique, one-minute high frequency data, and a substantial number of artificial agents, which enables us to develop of a wider variety of trading rules. The presence of 10,000 artificial agents in our experiment results in increased forecasting model stability and lower sensitivity to random factors. To summarise, the contributions of this study are as follows. Firstly, we investigate the efficiency of currency markets by analysing the profitability of technical trading rules at the frequency at which this trading actually takes place in the real world. Secondly, we take into account all the issues identified in the literature as potentially affecting the reliability of trading results and inference based on them: transaction costs, allowing agents to learn from their experience, evaluating the profitability of rules based on their predictive power rather than in-sample fit, and avoiding data-snooping biases by allowing all potential rules and their combinations to be traded on and evaluated by agents. Thirdly, we are the first to apply the STGP technique to analyse the impact of HFT on market quality, taking into account the market structure. The remainder of this paper is organised as follows: Section 2 presents the background and a review of the literature in the field; Section 3 discusses the experimental design; Section 4 presents the forecasting methodology used in the analysis; Section 5 discusses the empirical results and the paper concludes and suggests avenues for further research in Section 6.

Due to its recent emergence the HFT discussion is not yet supported by a great deal of solid academic research (Chordia et al., 2013). This paper contributes to the discussion by providing appropriate empirical evidence from the FX market about the profitability of HFT strategies based on one-minute historical data and their implications for market quality. We develop an intraday technical trading strategy using the STGP form of Genetic Programming for six of the most traded currency pairs and find evidence of HFT predictability and profitability after taking into account appropriate transaction costs. The STGP technique outperformed ex-ante traditional econometric forecasting models. The ability to observe all trading activities in our experiments enabled the investigation of the impact of HFT bid and ask orders on market quality. We find evidence that HFT enhances the efficiency of prices and play a positive role in the price discovery process by trading in the direction of permanent price changes and in the opposite direction to transitory pricing errors. However, the fact that HFTask orders are positively associated with pricing errors could have implications for manipulative trading strategies, inappropriate risk management practices or adverse order selection. We think that further investigation is needed to examine the specifics reasons for this positive association. The debate as to whether HFT is beneficial or harmful to market efficiency is likely to continue long into the future and there are a number of important moral and practical issues involved. Hansbrouck and Saar (2013) found that some HFT algorithms need only 2–3 ms to identify the arrival of an order, analyse it, and generate an order. This very high operational speed prevents human traders from appropriately observing the limit order book, indicating that market dynamics might be dominated entirely by the interplay between trading algorithms. Initially, there is the possibility of mistakes or unforeseen problems in the complex trading algorithms driving HFT. It is an open question as to whether HFT firms have the appropriate mechanisms in place for signing off their complex trading algorithms and allowing proper engagement of senior management in this. Unfortunately regulatory bodies do not have the resources or experience to investigate the code in HFT algorithms creating conditions for moral hazard. Secondly, substantial funds are being spent on computer systems and very high-speed data connections that may have little social value. For example in 2010, more than $300 million were spent on 800 miles of fibre optic cable laid between the Chicago Mercantile Exchange and the NYSE in order to shave three milliseconds off of trading times (Gennet, 2012). Thirdly HFT may be severely disadvantaging some investors. HFT imposes difficulties for investors such as pension funds seeking to purchase large blocks of financial instruments, because HFT software can detect and front-runs the order. Individuals and even large traditional investors have restricted access to the same type of trading equipment as HFT. HFT could potentially create negative externalities on other market participants due to continuous generation of submissions and cancellations of limit orders that increases the need for costly equipment updates and worsens market regulation (Gai et al., 2012). Nowadays, investors are cautious about the possibility of having their traders detected and headed off by high frequency traders. This is the reason why traditional investors allocate their orders in ‘dark pools’. ‘Dark pools’ are off-exchange trading platforms administrated by brokers where financial instruments are executed anonymously and the prices are not announced in advance. According to the recent UK Foresight report (2012), two-thirds of investors went into ‘dark pool’ transactions, tripling their market share in the last few years to 3.3% of the total trading volume. Trading in ‘dark pools’ is having several negative consequences, such as: increased expenses which affect the transparency of the market by imposing price obstacles for the other investors. The ASIC report (2013) suggests that HFT performed in off-market ‘dark pools’ are adversely affecting the quality of asset price information and widening the bid-ask spread for several assets. Finally, financial markets exist for the benefit of society by allowing funds to be raised and directed to the most productive ends. Society can benefit only when the market reveals the true value of an investment though its market price. However, HFT algorithms do not explicitly take into account the intrinsic value of an investment, they are only concerned about what happens to the price in the next few seconds. This short term approach may limit the social benefits of HFT. Even if in the short term HFT seems to increase market efficiency, the long term effects on ability of markets to allocate funds are unclear. Our empirical findings are one step towards a better understanding of the underlying principles of HFT and its implications for market structure and performance although given the complexity and importance of this area much more research is needed. There are several interesting research directions to pursue. It would be very interesting to use data collected at very short intervals even down to ones measured in milliseconds to get a better feel for the conditioned faced by the faster HFT. Another research agenda is to investigate whether HFTs have different impacts on different trader populations and types of trader by constructing markets composed of different number and types of artificial traders. This can be investigated in a systematic way looking at the numerous potential differences between traders (e.g. wealth, intelligence, speed of processing news etc.) and trader populations.