A note on the relationship between high-frequency trading and latency arbitrage

https://doi.org/10.1016/j.irfa.2016.06.014Get rights and content

Highlights

  • We simulate real-life trading at the millisecond interval.

  • HFT scalpers calculate NASDAQ NBBO in 1.5 ms.

  • HFT scalpers generate latency arbitrage opportunities.

  • Market efficiency is negatively affected by HFT scalpers.

  • We propose batch auctions in every 70 ms.

Abstract

We develop three artificial stock markets populated with two types of market participants — HFT scalpers and aggressive high frequency traders (HFTrs). We simulate real-life trading at the millisecond interval by applying Strongly Typed Genetic Programming (STGP) to real-time data from Cisco Systems, Intel and Microsoft. We observe that HFT scalpers are able to calculate NASDAQ NBBO (National Best Bid and Offer) at least 1.5 ms ahead of the NASDAQ SIP (Security Information Processor), resulting in a large number of latency arbitrage opportunities. We also demonstrate that market efficiency is negatively affected by the latency arbitrage activity of HFT scalpers, with no countervailing benefit in volatility or any other measured variable. To improve market quality, and eliminate the socially wasteful arms race for speed, we propose batch auctions in every 70 ms of trading.

Introduction

Wissner-Gross and Freer (2010) suggest that the time light travels between antipodal points on the surface of the Earth takes 67 ms, while recent computational advances transform HFT latencies below 500 μs (Bhupathi, 2010). Many HFT strategies are designed to exploit advantages in latency — the time it takes to access and respond to market information (Wah & Wellman, 2013). Schneider (2012) estimates that trading on latency advantages account for $21 billion profit each year. HFTrs are able to obtain such speed advantages over institutional investors by developing sophisticated trading algorithms combined with co-located computer systems, directly linked with trading venues. At the same time, market structure issues due to speed competition among HFTrs create the unintended consequence of allowing faster traders to gain revenue from trading with slower traders (McInish & Upson, 2013). The practice of HFT has generated several public controversies regarding its transparency and the fairness of market operations, as well as its implications for market quality (Wah & Wellman, 2013).

However, most of the empirical work on the topic lacks the ability to identify which trades and quotes come from HFT, making it difficult to examine how HFT affects the market and other market participants (Egginton et al., 2012, Goldstein et al., 2014, Hirschey, 2013). This is due to the fact that no publicly available dataset, including NASDAQ 120, allows researchers to directly identify all HFT (Baron, Brogaard, & Kirilenko, 2012). Egginton et al. (2012) argue that is hardly possible to identify orders generated by computer algorithms in the U.S. equities markets, with all previous studies using proxies to measure the level of algorithmic trading and HFT.1 The huge number of variables and very complicated cause-effect relationships among these variables and potential outcomes imposes another research obstacle (Felker, Mazalov, & Watt, 2014). Furthermore, empirically measuring informational differences between different investors represents a difficult task as investors' information sets are unobservable (Ding, Hanna, & Hendershott, 2014).

In contrast, this study uses a special adaptive form of Strongly Typed Genetic Programming (STGP) and real-time millisecond data from Cisco Systems, Intel and Microsoft to demonstrate the process of latency arbitrage in HFT. The STGP (described in Appendix C) is a sophisticated and extremely suitable trading algorithm that successfully replicates HFT scalping strategies. Wah and Wellman (2013) argue that questions about HFT implications are inherently computational in nature due to the fact that the speed of trading reveals details of internal market activities and the structure of communication channels. We subscribe directly to NASDAQ's Security Information System (SIP), which is called the Unlisted Trading Privileges Quote Data Feed in order to reproduce the HFT scalping strategies in an artificial stock market environment. Here, the impact of these strategies can be examined and new regulations evaluated to maintain the overall health of the financial system. Using STGP, we replicate the interactions between HFT scalpers and aggressive HFTrs and compare their performance under the same underlying trading order streams. In other words, we simulate real-life trading sessions, which allow us to avoid the obstacles in the studies discussed above. HFT scalping strategies originated as relatively simple spread detecting tools that came to understand the order book depth, posted on the best bid/ask and then moved quickly to the other side (Patterson, 2012). These straightforward flipping strategies evolved over time to become the modern HFT scalping strategies that nowadays dominate electronic exchanges, gaining favourable queue positions and generating a huge amount of cancelled orders. The aim of HFT scalping strategies is to gain a favourable queue position — any particular scalping strategy must have a high probability of entering the trade and an equally high probability of either exiting for spread or, if the spread cannot be gained, of immediately exiting in order to avoid losses (Bodek, 2013).

To summarise, the contribution of this study is three-fold. First, this is the first study to use an innovative trading algorithm and real-time millisecond data to provide empirical evidence of how HFT scalping strategies are able to calculate NASDAQ NBBO (National Best Bid and Offer) at least 1.5 ms ahead of the NASDAQ SIP, creating a large number of latency arbitrage opportunities.

The Securities and Exchange Commission (SEC) developed the Regulation National Market System (Reg NMS) in 2007 in order to protect fair access to the best stock price for traditional investors. According to Reg NMS rules, trading venues are required to provide trading messages to the primary exchanges such as NASDAQ and NYSE. The SIPs for NASDAQ, which are called Unlisted Trading Privileges Quote Data Feed, along with NYSE's Consolidated Quotation System, collect all relevant data and calculate the respective NBBO. Consequently, stock brokers are required to execute trading orders at NBBO prices or better (Ding et al., 2014). However, considering trading order information from all exchanges, the SIPs take some finite time, let us say δ milliseconds, to calculate and for the NBBO to be distributed. Computationally sophisticated traders equipped with front-running scalping strategies such as HFT scalpers can process the order flow in less than δ milliseconds and out- compute the SIP to calculate the NBBO. Under trading conditions of superhuman speed, quotes within an exchange could update faster than the exchange is able to distribute its new prices to other trading venues for NBBO evaluation. Our experiment detects the processing and calculation of both best bid and ask orders; it does this by simulating the communication patterns between HFT scalpers, aggressive HFTrs, NASDAQ SIPs and NASDAQ NBBO. Our empirical results demonstrate that the ability of HFT scalpers to create latency arbitrage opportunities makes trading more difficult and more costly for those traditional investors who lack access to sophisticated trading platforms.

Second, this study provides the first real-life trading evidence whereby direct access to exchanges and appropriate trading software could generate profitable opportunities for HFT companies. We demonstrate that HFTrs equipped with scalping trading mechanisms are capable of capturing substantial risk-free profits at the expense of institutional investors. We also measure the precise level of profits generated by HFT scalpers and the exact costs of latency arbitrage for other market participants. Our findings suggest that there is an arms race in speed and how fast market participants have to be to capture profit opportunities. The size of the arbitrage opportunity, and hence the harm to institutional investors, may depend on the magnitude of speed and the cost of cutting-edge speed improvements.

Third, we provide clear evidence of the implications of latency arbitrage for market quality and the relationship between market fragmentation and latency arbitrage strategies.

Our empirical findings indicate that latency arbitrage not only reduces the profits of other market participants, but harms market efficiency. We observe that HFT scalpers' latency arbitrage activity has negative implications on market efficiency as intraday volatility increases and market depth decreases. We propose an alternative financial market mechanism such as a batch auctioned market, which successfully eliminates latency arbitrage opportunities and improves efficiency. We suggest the implementation of batch auctions once every 70 ms, that is 334,285 times per 6.5-h trading day for each financial instrument. If trading orders are bunched together every 70 ms, HFT scalpers could face a queuing risk leading to a less harmful market quality effect.

Our study is particularly timely as policymakers around the world are still debating as to whether HFT is beneficial or harmful to market efficiency (Manahov, Hunson, & Gebka, 2014). To a certain extent this study can be seen as a tool assisting regulators in the more rigorous evaluation of the financial market.

The remainder of this paper is organised as follows: Section 2 comprises the literature review on the topic. Section 3 presents the experimental design of the three artificial stock markets and data description. In Section 4, we examine the HFT scalpers' latency arbitrage activity and profitability and investigate the implications of HFT scalpers on market quality and the associated regulatory measures. Finally, Section 5 concludes the paper. Additional clarifying and technical material can be found in Appendix B.

Section snippets

Related literature

While Ready (1999) and Stoll and Schenzler (2006) perform empirical analysis to show how slow traders' orders provide a free trading option for fast traders, Cohen and Szpruch (2012) consider a single asset market model of latency arbitrage with one limit order book and two traders possessing different speeds of trade execution. This is to demonstrate that the fast trader employs a front-running strategy to capture the quantity that the slower trader intends to trade and generate a risk-free

Experimental design

Due to advances in technology and the rapid growth of high frequency trading, financial markets have eliminated human intermediation in the trading process and replaced them with electronic limit order books, which have led to the growth of trading algorithms as one of the main investment tools. Some of the trading algorithms generated imitate the behaviour of humans in the trading process, while over the last few years, these trading algorithms have substantially improved their speed to match

High-frequency trading and latency arbitrage

First, we examine what happens to the trading orders of Cisco Systems, Intel and Microsoft after being submitted to the three artificial stock markets.3 Jarnecic and Snape (2014) examine the order submission strategies of HFTs and traditional traders in the limit order book and observe that high-frequency participants cancel orders of all durations

Conclusions

The application of sophisticated computational trading strategies at very low latency has increased over time. Significant trading software improvements are constantly introduced, raising operating costs and increasing competitive advantage among market participants. As communication and trading speed in financial markets has decreased over time, regulators face additional challenges in terms of addressing the speed differentials of market participants.

In this study, we simulate real-life

References (55)

  • B. Biais et al.

    Equilibrium high frequency trading

  • H. Bodek

    The problem of HFT. Collected writings on high frequency trading and stock market structure reform

    (2013)
  • J. Brogaard

    High frequency trading and market quality

  • J. Brogaard et al.

    High frequency trading and the execution costs of institutional investors

    The Financial Review

    (2014)
  • J. Brogaard et al.

    High frequency trading and price discovery

    The Review of Financial Studies

    (2013)
  • E. Budish et al.

    High frequency trading arms race: Frequent batch auctions as a market design response

  • A. Chakraborti et al.

    Econophysics review: agent-based models

    Quantitative Finance

    (2011)
  • S.N. Cohen et al.

    A limit order book model for latency arbitrage

    Mathematical Financial Economics

    (2012)
  • S. Ding et al.

    How slow is the NBBO? A comparison with direct exchange feeds

    The Financial Review

    (2014)
  • D. Easley et al.

    Differential access to price information in financial markets

  • J. Egginton et al.

    Quote stuffing

  • T. Foucault et al.

    News trading and speed

  • M.A. Goldstein et al.

    Computerized and high-frequency trading

    The Financial Review

    (2014)
  • J. Harris et al.

    The sound of silence

    The Financial Review

    (2014)
  • T. Haynes et al.

    Strongly typed genetic programming in evolving cooperation strategies

  • T.D. Haynes et al.

    Type inheritance in Strongly Typed Genetic Programming

  • T. Hendershott et al.

    High frequency trading and price discovery

  • Cited by (0)

    View full text