The Gateway to Algorithmic and Automated Trading

High frequency trading: Reaching the limits

Published in Automated Trader Magazine Issue 41 Q4 2016

The tremendous growth in high frequency trading (HFT) seems to have reached its limits in recent years. Massively increased costs for infrastructure and relentless competition are probably to blame.

Dr Orçun Kaya


Dr Orçun Kaya is an economist in the banking, financial markets and regulation team of Deutsche Bank Research. He focuses on capital markets, corporate financing as well as financial markets infrastructure. Previously, he was a research assistant and coordinator of the Masters program in Money and Finance at Goethe University Frankfurt, where he gained his PhD.

Advances in technology in recent years have made it possible to place buy and sell orders at an ever increasing pace, to automate older, existing strategies and to create new, previously infeasible ones. Trades can now be executed in a few microseconds, one hundred thousand times faster than the blink of an eye (normally about 100 to 400 milliseconds).

As a result, high frequency trading (HFT) - the use of sophisticated computer programs of extraordinarily high speed for submissions and cancellations of orders to realise small profits - has become widespread, with equity markets being one of the areas of focus. HFT's market share has boomed over the last ten years or so and electronic market making has become an integral part of securities trading.

In contrast to its early rapid expansion though, the HFT share as a fraction of total equity trading has been declining since the financial crisis (see Figure 01). An assessment of potential factors that hobble market growth can shed some light on the future growth potential of the HFT industry.

Figure 01: Declining HFT market share in equity trading

Figure 01: Declining HFT market share in equity trading

High frequency trading in a nutshell

Computing technology has revolutionised the way financial assets are traded. Handling orders without human intervention, where computer algorithms automatically make trading decisions, submit orders and manage positions has become ingrained in financial markets. Even though algorithmic or automated trading (AT) has existed for several decades, the speed at which it is conducted has increased tremendously over the past ten years. Specifically, HFT is one area within automated trading in which trade execution occurs at particularly high speeds.

Even though HFT is considered a subset of AT, it has important additional features. First and foremost, trade execution times are usually measured in microseconds for HFT and a large number of order submissions and cancellations are required with the arrival of new market information. High frequency traders have very short holding periods and are usually neutral or have no positions at the end of a trading day. Across the HFT landscape, profit opportunities are extremely time-sensitive and low latency times for trade execution are of central importance to realise at least small profits. In this vein, HFT firms heavily invest in high speed connections to exchange facilities and usually place their trading engines close to exchange servers either via co-location and/or proximity services. Ease of trading and the liquidity of the financial product are of particular importance in applying HFT strategies. As a result, high frequency traders focus only on very liquid assets like blue chip stocks and typically do not trade less liquid instruments, particularly those that are traded over-the-counter such as corporate bonds or exotic derivatives. In constantly providing liquidity, high frequency traders essentially take on the traditional role of a market maker or, perhaps more accurately, that of a liquidity provider. Unlike officially designated market makers, HFTs have no obligation to remain in the market (particularly during periods of high volatility) and may withdraw at their discretion. This has led critics to conclude that HFTs act as market makers only on their own terms without any broader obligation to other market participants.

Figure 02: US HFT revenues

Figure 02: US HFT revenues

Recent trends in HFT

Briefly, the evolution of HFT can be divided into two episodes. The pre-crisis period was marked by the rise of HFT on both sides of the Atlantic. As Figure 01 shows, in Europe the share of HFT in total equity trading surged virtually from zero in 2005 to around 40% in 2010. In the US, 20% of the trades already came from HFT in 2005, which peaked at 60% of the market in 2009. However, with the financial crisis, the rise of HFT came to a halt and market share began to recede. As of 2014, the share of HFT in equity markets had come down to 35% and 50% of the total market in Europe and the US, respectively. Three factors explain the decreasing HFT market share in total equity trading:

  • a decline in revenues and profits due to the increasing cost of HFT infrastructure
  • relentless competition within the industry
  • the rise of alternative trading platforms.
  • Forthcoming regulation and potential taxation will also act as drags on the industry.

Declining revenues and profits

Overall, HFT firms' revenues in the US have slumped from about 7.2 billion USD in 2009 to 1.3 billion USD in 2014 (see Figure 02). This drop may have been partly due to fierce competition which has pushed smaller contenders out of the market and led to trading opportunities disappearing more quickly. Indeed, HFTs execute a huge number of trades with a very small profit per trade. More specifically, HFTs' average profit per share traded is a fraction of a cent and in order to generate meaningful profits, the HFT population in its entirety has to execute millions of trades every day. Also relevant in this context is the role of price volatility. In recent years, episodes of heightened volatility have become common in financial markets. For example, between 2010 and 2015 very rare events (a five standard deviation move in asset prices) occurred 2.5 times more often than between 2001 and 2006 in US equity markets. Volatility in financial markets favours HFT, at least in principle. However, the detrimental impact from increased competition may have prevailed over this positive effect: in a market with more advanced technology and infrastructure, price disparities across venues will be spotted and eliminated ever more quickly. Hence, in the past few years, desirable trading opportunities have probably become scarcer even for the fastest traders.

Moreover, increased costs also push hard on the HFT industry. High frequency traders maximise their revenues via their absolute and relative speed in executing trades. In this vein, co-location services that give access to market information ahead of other market participants, even if only by a few microseconds, are of central importance for latency-sensitive traders. Spotting the growing demand for close-proximity servers, however, major exchanges doubled or tripled their co-location service fees over the 2010-15 period. Other infrastructure providers that specialise in laying cables to transfer data at high speed between electronic trading venues have entered the market as well. HFT firms are, de facto, compelled to use and pay for the most advanced infrastructure. For HFTs, speed matters in the relative sense as high frequency traders have to be at least as fast as (or faster than) their competitors. All in all, the cost of high speed trading infrastructure has risen significantly in recent years.

The bottom line is that HFTs have come under pressure on both the revenue as well as the cost front. Unsurprisingly, then, even though a lack of available data prevents a comparison of the number of competing HFT firms over time, data for the US shows that the average profit per share traded has halved, from a tenth of a penny in 2009 to a twentieth of a penny in 2015.

The role of dark pools

Another challenge that HFT firms face is the advancement of Alternative Trading Systems (ATSs). In short, ATSs are trading venues other than exchanges. Among the most important ATSs are 'dark pools' that generally match buyers and sellers anonymously, without public information on the best prices. Dark pools only publicly report trade volumes and prices after a trade has been executed. Put differently, providing pre-trade anonymity can be seen as the core business of dark pools. The main aim is to have a smaller impact on prices compared to trades carried out on exchanges, particularly for relatively large trades.

ATSs have been around since about 1960 and so are not a new invention. However, ATSs in general and dark pools specifically benefited from the adoption of a new regulatory framework ('Regulation of Alternative Trading Systems', or RegATS) by the SEC in 1998. As a result, the number of dark pools has grown significantly in recent years and there are now about 50 dark pools operating in the US and 15 in Europe. The greatest catalyst for the rise of dark pools is institutional investors' growing desire to trade large amounts while minimizing market impact. Even though this has always been a requirement of institutional investors, it has become acute with the advance of algorithmic trading and HFT. With the active enrolment of those traders in financial markets, order sizes have diminished and asset prices move more quickly in one direction or the other. Within a dark pool, some of this can be avoided and counterparties can trade securities as a large block of shares in one transaction.

Figure 03: Equity trades in ATS in the US

Figure 03: Equity trades in ATS in the US

A lack of available data prevents a detailed analysis of dark pool trading, but trading in ATSs in general serves as a good proxy for general trends. In recent years, ATSs have attracted significant trading volumes, and an increasing share of trades are now executed away from the exchanges. In the US, the number of shares traded on ATSs jumped from 57 billion in 2014-Q3 to 70 billion in 2015-Q4 (see Figure 03). Other estimates are that around 40% of US trades took place off-exchange in 2014, up from a mere 16% in 2008. That said, HFTs are not per se locked out of these venues and in principle have access to ATSs. However, they are mostly active on non-block oriented pools (see 'Types of Dark Pools' below), i.e. ATSs that focus on matching small trades, while they are less engaged in block-oriented dark pools where certain minimum requirements exist regarding order size. In other words, the growing role of dark pool trading limits the ability of high frequency traders to apply their strategies to very large block trades, which traditionally provided good opportunities for the HFT community. This is likely a source of significant further pressure on the revenues of HFT firms.

Types of Dark pools

Block-oriented dark pools
These venues specialize in matching block trades. They are particularly suitable for institutional investors such as asset managers and pension funds. An order is considered a block trade if it consists of a very large number of shares or a high market value, i.e. the hurdle could be a minimum of 10,000 shares or a total market value of 200,000 USD. It is almost compulsory to trade such orders in dark pools as they might exceed the average daily volume. That said, for very large trades it might sometimes be difficult to find a match even in a block-oriented dark pool due to the size of the order.

Non block-oriented dark pools
On these venues, there is no minimum order size required for access. Nowadays, the majority of dark liquidity is traded in these venues (as opposed to those catering only to block trades).
Non block-oriented dark pools tend to trade more in retail-sized orders and allow participants to trade small quantities. As a result, unlike block-oriented pools, HFTs are actively involved in these venues. The enrolment of HFTs results in favorable liquidity characteristics and improves the likelihood of a timely match for trading counterparties. The risk is, of course, information leakage.

Stronger regulation is on the horizon

Apart from the market dynamics themselves, the forthcoming tougher oversight and regulation will drastically shape the future of HFT as well. After the 'Flash Crash' in May 2010, when major US indices dropped around 10% intraday, policy makers have been taking a deliberate stance against HFT both in the US and in Europe. Indeed, in the eyes of some observers, algorithms do not follow economic fundamentals and merely chase mechanical patterns which may significantly harm the functioning of financial markets. To mitigate these stability concerns, the Markets in Financial Instruments Directive 2 (MiFID 2) introduces major changes to how financial instruments are traded and distributed in Europe. It imposes stricter requirements on trading venues and market participants that engage in AT and, specifically, HFT. Among these are order-to-trade ratios that prevent overly rapid and frequent submitting and cancelling of orders as well as systems and risk controls to ensure resilience of venues. Also included in MiFID 2 are direct market access limits and algorithm disclosure requirements for HFT firms. In Europe, MiFID 2 rules are expected to come into force in January 2018. In the US on the other hand, authorities have been somewhat slower in addressing HFT regulation. The SEC aims to:

  • introduce an anti-disruptive trading rule
  • improve risk management practices for trading algorithms
  • enforce stricter use of its core tool of registration and oversight (see Automated Trader, issue 39, page 28-29).

Both the CFTC and SEC rules for HFT are expected to be finalised in early 2017.

Furthermore, some exchanges such as IEX (and more recently CSX) are trying to lure non-high frequency traders to their platforms through the introduction of 'speed bumps' which aim to reduce the speed advantages enjoyed by some traders who have made big investments in their infrastructure.

Closing remarks

As the HFT technology matures, the costs of high speed trading have increased and fierce competition for speed seems to have reached its limits. As a result, HFT firms are seeing both their revenues and profits erode. Restricted access to dark pools' block trades has been challenging HFT firms' trading strategies further. On top of this, forthcoming stricter prudential regulatory oversight may result in an overhang of capacity in the HFT industry. All in all, the glory days of HFT seem to be over and HFT will probably offer reduced opportunities in developed markets in the future.

An earlier version of this article has appeared as a Deutsche Bank Research Publication in May 2016.


Brogaard, J. 2010 High frequency trading and its impact on market quality (Working paper)

Chlistalla, M. 2011 High frequency trading (Deutsche Bank Research, research briefing)

Gomber, P., Arndt, B., Lutat. M. and Uhle, T. 2011 High frequency trading (Working paper)

Shorter, G. and Miller, R. S. 2014 Dark pools in equity trading: policy concerns and recent developments (Congressional research service report)

Vaananen, J. 2014 Dark pools and high frequency trading for dummies (1st edition, John Wiley & Sons)

Kaya, O. 2016 High frequency trading: reaching the limits (Deutsche Bank Research)