The focus on execution latency is rapidly giving way to decision latency (the pre-trade time to recognise a trading signal and place or cancel an order) and value latency (the time to confirm and process the transaction flows post-trade and extract their impact on alpha). The obsession with time and speed has become pervasive. Risk too has won mind share, perhaps not surprisingly given the collateral damage sustained by some high-speed broker-dealers. Attitudes towards regulation are fragmenting. While many feel the rule making process has gone astray, trading firms are resigned to coping with it and are open to changing priorities, strategies or even jurisdictions in order to remain viable.
Key points to emerge from the data
The survey recorded responses from 615 market participants and professionals. The number of respondents for Asia and emerging markets increased by 50% from the previous survey, with many more buy and sell side firms featuring in all categories. Overall, 41% of respondents were buy side, 20% sell side, with the remainder from exchanges, market infrastructure and technology firms. One regulator and a few academics also responded.
Both big and small players were well represented, with buy side respondents from traditional asset management firms, hedge funds, proprietary trading firms, family offices and trading arcades all responding in good numbers. Of the traditional asset managers, almost 16% had over $10 billion under management, whilst one-third of hedge funds managed more than $100 million and 17% managed over $1 billion. On the sell side, over 20% had balance sheets in excess of $500 billion, while 43% had more than $5 billion in assets and liabilities.
Top Trading Challenges
The survey posed 144 probing questions to record the most panoramic picture of global trading trends ever. In total, the survey produced 67% more data than last year. This preliminary analysis highlights just a few of the many conclusions that will be exhaustively mined in the final report.
Key Challenges and the Latency Squeeze
While two-thirds of the buy side were still focused on finding alpha, more were concerned with regulation (31%) than previously. Managing real-time risk (36%) and the need to re-engineer models (37%) were also listed amongst the top challenges.
A clear majority of the sell side were focused on real-time risks, whilst keeping up with regulation, finding alpha and coping with the speed and volume of data feeds all registering as significant concerns (43%, 41% and 29% respectively).
Latency issues were another big worry, of course, with a majority of sell side, market infrastructure and tech firms confirming that their latency focus had grown over the previous two years, often strongly. Smaller broker-dealers now seem to be seriously caught up in the race to zero, with almost 80% expressing high concerns. This matches the concern shown by the large investment banks and market infrastructure companies. Hedge funds and prop traders are not far behind, with nearly two-thirds confirming a continuing or growing focus on latency issues.
Changing Focus on Latency
Getting (and keeping) an edge
Large investment banks are investing in faster circuits and new technology. Smaller sell side firms are now actively rolling out colocation schemes. Traditional asset managers are still relying on best-of-breed brokers, while around a quarter are exploring colocation and exotic hardware such as FPGA and GPU. Hedge funds and prop traders are prioritising colocation and faster inbound data circuits, especially with FPGA and GPU accelerators. Meanwhile, investor-traders - i.e. family offices, trading arcades and day traders - are also pursuing colocation, finding better brokers or switching to DMA channels. There seems to be a levelling down of latency going on here, with all the specialist HFT technologies beginning to go mainstream.
Adoption of the more exotic hardware solutions continues to increase, with nearly 9% of all firms already using niche hardware solutions and over a quarter expecting to use them within three years. While usage remains modest at between 5% and 10% for buy and sell side firms across all regions, this is expected to climb to around 20%+ within three years. The real growth, however, is found amongst market infrastructure and tech firms. Here, current usage of FPGA was confirmed at already over 20% and GPU over 15%. Within another three years this usage is expected to grow to 35% and 24% respectively, making these superchargers available to the masses.
Success factors for low latency
Indeed, cloud technology and colocation appear to be growing partly because of these specialist capabilities. For example, nearly 8% of buy side firms cited FPGA or GPU as their reason for using cloud. Meanwhile, between 20% and 25% of buy or sell side colocation users cited reduced costs and increased access to third party services as the key drivers of their hosting decisions. Additionally, nearly two-thirds colocate to access market data. While this year's forecasts for in-house usage of exotic hardware are perhaps more modest than last year (up to 40% penetration in three years predicted last year versus 25% to 35% this year) they continue to show solid growth, and the actual shift to third party services fully substantiates the earlier optimism.
Other new technologies that made a notable appearance in this year's data were microwave telecommunications and atomic clocks. So far, use of microwave has reached 3% to 5% amongst buy and sell side firms, but this is expected to more than double by next year. Indeed, 16% of all respondents see microwave as offering the greatest gains in the low latency game. Again, third-party service providers have the most ambitious plans, with 15% penetration expected by next year, although microwave's limited capacity may moderate these ambitions.
Cloud computing shows a similar profile, with 36% of third-party service providers using it, followed by buy side firms (26%) and sell side firms (10%). Most people are looking for cost savings or flexibility, although increasingly it will just be another embedded technology in a third party offering.
Use of atomic clocks and hardware latency measurement devices has now reached significant levels; almost one-third of service providers have already deployed these technologies, and nearly one in four sell side and one in eight buy side firms are currently using them. Steady growth is predicted over the next two to three years, especially for the buy side.
Virtually all respondents globally described the latency characteristics of their infrastructure as having improved, and this is expected to continue. For example, over 50% of global investment banks and technology firms expect within two years to be near the leading pack, if not at the absolute front themselves, in terms of time-to-market. For market infrastructure companies, nearly 70% expected to be market leaders in latency. Even 20% of traditional asset managers, who mostly rely on their brokers for speed, expect to be near the leading pack by then, along with 40% of hedge funds and prop traders. Clearly, they all have expectations that competition will perhaps erode over time. However, technology itself will narrow the range of relative performance advantage, driving ever more firms to look beyond execution latency for their competitive advantage.
Technology Now and to Come
When the pace of change becomes a blur, quant traders apply software analytics to the problem. Consequently, use of transaction cost analysis (TCA) is now growing across the board. For the sell side, hedge funds and prop traders, TCA usage is expected to double in the next year. For traditional asset managers growth is more modest, although TCA was already much better established amongst this sector, with the majority of traditional asset managers already using TCA for assessing both brokers and their own internal traders. For the sell side, it will soon become mainstream practice as well. Similarly, execution consulting looks set to gain traction from the relatively low numbers, with around 13% of traditional asset managers and just under 4% of hedge funds currently making use of this service, but a further 13% of asset managers and almost 11% of hedge funds now considering using the service.
Algos for everything mean new latencies to measure
While the obsession with latency continues, differentiation through other means has become the key focus for many firms, and here the survey data provided some very useful pointers. Execution algorithms on the sell side and systematic algorithms on the buy side are now mainstream, with 60%+ of firms (both soon rising to 70%) using them. The majority of buy side firms are also deploying their own execution algorithms as well, while a majority of sell side firms now use systematic algorithms, for example for market making or to offload risk trades on their own books. Other complementary algorithms for portfolio management, risk management pre-trade checks, post-trade processing, hedging, performance monitoring etcetera are now spreading, and soon the majority of both buy and sell side firms will be using them. Last year's conclusion that algorithms were rapidly becoming end-to-end was thus reconfirmed, with strong growth continuing.
Indeed, among buy side firms that are struggling with latency and volume issues, nearly 60% say that almost all of their executable orders are machine-generated without any manual intervention. The continued increase in low- or no-touch flow confirms the end-to-end technology focus and the growing importance of decision latency.
Participants were asked to provide an estimate of the decision latency between capturing market data off the wire and sending out an order to the execution engine. With responses ranging from single digit microseconds all the way through to seconds, decision latency clearly has huge scope to become the next major development focus for many firms.
The interest in post-trade value latency was similarly evidenced by several developments, including the expansion in automated hedging, post-trade algorithms and risk management. The whole post-trade cycle is already dynamic for many firms, and is steadily moving from overnight batch into real time for the remainder.
Unconventional Data for Systematic Trading
Some insight into value latency can be gained by looking at mean holding times for assets in a systematic trading algorithm. Over 40% of the buy side estimated asset holding times of seconds to hours. With fungible products and a central counterparty on a single platform this may be relatively easy to confirm, but traders are now trading across venues, asset classes and jurisdictions, so the race to zero for value latency is now well underway.
The survey confirmed that trading on a single platform is becoming less viable for statistical arbitrage, spread trading or relative value strategies. For both buy and sell sides, around half the firms using these strategies have to work cross-market.
Risk is another area where traders are trying to juggle multi-market event flows. Around two-thirds of the buy side or over 80% on the sell side now expect to be able to detect and react to a risk 'problem' within seconds or possibly a minute or two at most. This is perhaps not surprising, given a trading environment where companies can go bust in minutes or even seconds. However, the survey documented the enormous range of signals and event streams that need to be parsed to recognise a rogue algorithm or abnormal market. For virtually all algorithmic traders, real-time risk monitoring is now essential, ideally with automated feedback loops to the trading engines.
Big Data Big Time
Already, 10% of all buy side firms and 8% on the sell side use algorithms to monitor news feeds, and within three years this is expected to rise to 24% and 16% respectively. Among firms running systematic models, 15% of buy side firms and a huge 24% of sell side firms confirmed that they are carrying out their own semantic analysis of news, with lesser proportions using pre-cut sentiment metrics or news analytics. The highest buy side enthusiasm is found in Asia Pacific and Continental Europe (each over 20%), while North American sell-siders using news in their systematic models have reached 33%. This is remarkable, and even closing in on the numbers using latency metrics (27% of the buy side systematic traders overall, or 41% on the sell side).
Over 50% of market participants estimated that confirmed compliance requirements would cost them over 10% of their development budget, while another 45% of firms reckoned they spend a further 10% or more on possible, but still unconfirmed, regulatory measures. This is clearly a significant overhead and a competitive disadvantage for the main jurisdictions affected.
It should not be surprising that only 14% of market participants felt the current regulatory proposals to be well judged, while another 10% were sympathetic to the slow pace of rule-making, given the complexity of the issues. However, 36% were worried that the proposals still seemed unclear, and another 39% considered the pace of change to be too fast, or that the proposals would positively damage the market. Hedge funds, proprietary traders and smaller sell side firms were the least sympathetic, while traditional asset managers appeared to have the greatest approval of the regulators' intentions.
Attitudes to Regulatory Change
Despite some limited sympathy with the aims, in practice most market participants had little expectation of a positive result. Only 14% felt that politicians or regulators had a good understanding of the issues. Even traditional asset managers expected a poor result by a factor of more than two to one.
Only 13% of those expressing an opinion felt that market speeds had become too dangerous. Asked to state their own preferences regarding methods that might slow the markets, respondents' top three choices were maximum order-to-fill ratios set by the trading venue, minimum order resting times, also set by the venue, or minimum resting times set by a regulator. All of these were minority interests.
When asked to rank the same regulatory options by the damage they might do to markets, two-thirds of responses highlighted the financial transaction tax on all filled orders as causing the greatest harm, while nearly a quarter predicted that a minimum resting time would also be unhelpful.
However, when asked how they would cope if regulators were to go ahead with plans to slow the market, 34% of respondents said their most profitable trading strategy would probably not be materially affected. Another 19% felt they would be able to adapt, or indeed already had an alternative business strategy. Only around 5% thought it might risk the existence of their firm. Somehow they would muddle through, including a not-so-silent majority who said they might consider moving to another jurisdiction.
Expectations for how the overall market would cope were more mixed but less hopeful. While only 17% thought liquidity might actually improve, 45% believed that even if liquidity fell, the market would survive. Some 31%, however, deemed the proposals unworkable as multilateral approval would not be achieved and liquidity would simply move, while 27% thought liquidity would be severely damaged and spreads inevitably widen. Market sentiment is clearly both fractured and fractious. The full Global Trading Trends Survey report due for release in April will cover these and many more issues in detail, including:
• Where are the hot spots for automated traders?
• Is Asia set to overtake the established Western centres? Are any other emerging markets beginning to extend their reach by re-engineering their capital markets?
• How do attitudes to HFT vary by geography? Are regulators in Asia and elsewhere seen as intrusive as in the US and Europe?
• How are broker offerings changing to adapt to the changing landscape?
• How are buy side trading strategies and asset class portfolios adapting to these tectonic changes in the financial market landscape?
• If technology diversity is strengthening the hand of service providers, how does this impact firms' strategies for colocation or proximity data centres?
• How is software engineering for systematic and execution algorithms adapting to the new market ecology of speed and smart trading? Have the trends towards ever greater size, complexity and adaptive learning slowed or accelerated?
• What is big data really about? How is it impacting trading strategies and risk management?
• Which of last year's trends have really carried forward and developed?
• What are the trends to watch for next year?
To pre-order your copy of the 100+ page report, go to www.fa5t.net/1g0.