Posts

The rise and rise of artificial intelligence

Recent announcements from some of the largest banks show artificial intelligence (AI) working its way further into financial markets.

Credit Suisse has announced it is to deploy 150 new ‘robots’ over the course of the year, with an overall aim of cutting CHF 4.8 billion (GBP 3.7 billion).

UBS has unveiled a new AI system which uses machine learning to develop strategies for trading volatility on behalf of clients. The bank claims that this is the first ‘adaptive strategy’ product offered by an investment bank.

J.P. Morgan is developing a machine learning technology called LOXM which aims to improve execution quality in the bank’s European equities business. As the buy-side increasingly focuses on execution quality, this is driving ever greater adoption of algorithmic trading across asset classes. LOXM is programmed to learn from historical trading patterns and tweak its algorithmic strategies accordingly, using a technique J.P. Morgan calls ‘deep reinforcement learning’.

The ability to adapt and learn without human intervention allows LOXM to optimising the execution gains of algo trading.

Mosaic Smart Data is looking at how AI can improve trading across asset classes, taking on the challenge of providing machine learning capabilities to the FICC markets, which have far less standardised data and a greater portion of voice trading.

Mosaic provides both real time and predictive analytics insights for sell-side FICC traders, giving them a view of their market in a way that takes in far more data than a human being is able to comprehend. This augments the human trader’s capabilities and could lead to significant performance gains for sell-side FICC departments.

While initial uses of AI focused on process improvements, it is significant that the technology has reached a level where its insights are now helping to influence trading itself.

Although we are still some way from a fully automated robo-trader, this represents a significant increase in confidence in AI technology.

The potential benefits for corporates in algorithmic trading

Curtis Pfeiffer, Chief Business Officer at Pragma Securities, explains to FX-MM how corporates could stand to benefit from using algorithms for FX execution.

Why should corporates consider using algorithms for FX execution?

Corporations want to maximise profit, and a penny saved is a penny earned. Algorithmic trading can contribute to the bottom line by significantly reducing FX trading costs. Corporations trade on the order of $70 trillion a year – roughly the same as the total global GDP. On such large amounts, basis points matter.

That’s why, to fulfil their mission, corporate treasurers are increasingly focused on ensuring that they get best execution on their FX transactions, which includes using the best available trading tools and practices.

What advantages do algorithms have over other trading techniques?

With the speed at which trading is conducted today, the proliferation of trading venues, and sheer levels of information that is processed, it is simply impossible for a human trader to stay on top of all the data that the market is generating.

There are four core benefits to algo execution:

  • Breaking up a large order into multiple smaller pieces means, on average, paying less than trading in a block
  • Building algorithms on top of an aggregated liquidity pool effectively narrows the spreads being traded on
  • Building algorithms on top of an aggregated liquidity pool effectively narrows the spreads being traded on
  • Algorithms have the ability to provide liquidity as well as to take prices, allowing patient traders to capture part of the bid-offer spread
  • Automation frees treasurers and traders to focus more of their time on those issues where human intelligence and judgement add the most value.
What factors should investors consider when choosing an FX algorithm?

First, corporations should understand the bank’s liquidity model for their algorithmic offering – principal, agency or hybrid.

Bank algos access liquidity differently depending on the model. A pure principal algo accesses just the host bank’s liquidity, which also provides indirect access to other liquidity pools in the marketplace. Agency models do not interact with the host bank’s liquidity, but are able to provide liquidity on ECNs as well as taking prices, potentially capturing part of the bid-offer spread for the customer.

Hybrid models can offer the best of both worlds, though customers should understand how the bank manages its dual role as principal and agent. Corporations should assess the liquidity pool underlying each bank’s algorithms to determine which model will be most effective.

Second, corporations should be satisfied that their bank provider has first class algorithmic trading tools – either through a major investment it has made in algorithmic trading research and development internally, or by partnering with an algorithmic technology specialist. Smart algos have sophisticated order placement logic, change their behaviour based on pair and time-specific liquidity patterns, and make intelligent and dynamic use of the real-time liquidity available across venues – for example based on order fulfilment rates.

Provided liquidity and investment checks out, corporations can consider algorithmic trading as another service their banks provide, and direct flow as part of the overall banking relationship.

Finally, best practice is to use TCA after the fact to track performance across bank providers and make sure all is as expected.

To read more, please visit the FX-MM website here.

In search of FX liquidity

Foreign exchange (FX) is one of the world’s most liquid markets, with around USD 5 trillion exchanged across borders every day.

However, there is a perception in the market that liquidity is on the wane.

This is not necessarily true, according to David Puth, CEO of CLS. Speaking to Euromoney, he said “There is a tendency for market participants to believe that liquidity was better in the past. From what we see at CLS, liquidity appears to be very strong. It is, however, different, with liquidity widely dispersed over a number of different trading venues.”

The pessimism may in part be as a result of the increasing difficulty in defining exactly what liquidity means in the modern market, and measuring it accurately.

This was one of the questions which a recent report on liquidity in the Americas from the Bank of International Settlements (BIS) attempted to address.

Traditional liquidity metrics, such as cost metrics, quantity metrics and trade impact, have their uses, but the report finds that none are a perfect way to measure liquidity in the modern market.

This is important because one thing which is clear is that the modern FX market is becoming increasingly complex, making understanding liquidity more difficult.

The market, like many others, is fragmenting as electrification proliferates the number of trading venues and sell side participants put more emphasis on internalising trades.

Whether this fragmentation is having an impact on traders ability to trade, remains an open question.

The BIS report indicates that fragmentation does appear to be having some impact on liquidity measures, particularly when it comes to periods of market stress.

It gives examples such as the 2016 British EU referendum and flash crashes, where traditional liquidity metrics appear to have been impacted across a number of currency pairs, at least over the short term.

Dan Marcus, CEO of ParFX, points out that sometimes individual metrics don’t always give the full picture. “It may be the case that volumes are down from where they were… [However] on ParFX we do not see evidence of a problem with market depth or the ability for traders, who need to trade, fill orders.”

This is in part because, while technology is driving fragmentation, it is also creating opportunities to aggregate liquidity in more efficient ways.

“Buy-side traders have responded [to FX market fragmentation] by turning to algorithms and taking on more execution risk themselves”, says Pragma’s CEO David Mechner.

Liquidity is the lifeblood of the FX market, it is vital that the market can measure it in a way which gives an accurate representation of what it is like to trade. One solution, suggested by Mechner, is a consolidated tape, much like in equities. Until then, the market should think carefully about the metrics used to measure the market and ensure they are fit for purpose.

Algo trading on the rise as Pragma establishes European presence

The decision by Pragma to set up a base in London shows how the UK’s capital remains the natural hub for algorithmic currency trading despite the UK’s looming exit from the European Union.

While the debate about the future of London in a post-Brexit environment continues to rage on, there are many who continue to recognise the role of London at the centre of the USD5 trillion currency market.

Algorithmic trading in particular continues to rise in popularity. A report from Greenwich Associates found that the proportion of volume-weighted FX trading executed algorithmically has increased two and a half times in the past three years.

This trend was further highlighted by Pragma Securities, the multi-asset class provider of algorithmic solutions, which established a new connectivity presence in London to service its growing international client base.

London currently accounts for more than a third of all currency trading activity globally, according to the BIS. In a news article in FX Week, David Mechner, CEO of Pragma, expressed confidence in London and its role at the centre of European and international financial markets.

“Equinix’s LD6 site offers Pragma360 clients access to state-of-the art technology and the largest ecosystem for foreign exchange trading globally.

“The banks we service need state-of-the-art trading capabilities for their traders, and buy-side and corporate clients, making LD6 a natural fit.”

Pragma is not alone in its bullishness on London’s future, and it is clear that maintaining a data centre presence remains crucial to an institution’s trading operations, particularly for FX trading. The Financial Times recently reported on Dutch data centre operator Interxion’s £30m investment in its site in London’s Brick Lane.

Curtis Pfeiffer, Chief Business Officer at Pragma, also highlighted the benefits of proximity to London and risks of leaving London’s FX ecosystem.

“We are moving forward with this large capital expenditure because London, as the largest FX trading centre in the world, hosts the largest datacentre ecosystem for low-latency FX trading applications and we do not see that changing any time soon,” said Curtis.

“Institutions will be reluctant to leave the data centre ecosystem in London, which has increased in size significantly over the last 10 years as a result of a network effect – everyone wants their trading servers to be where everyone else’s are. By leaving that ecosystem, a firm could disadvantage themselves and their clients.”

Algo trading and interest in emerging market currencies will grow in 2017 driven by hunt for FX liquidity

Traders at across both buy and sell side are reporting that they plan to make more use of computer algorithms to trade FX in 2017 and are also setting their sights on traditionally less-traded currencies.

This matters. Foreign exchange – or FX – is the world’s largest and most liquid market, with around USD 5 trillion exchanged every day across borders.

FX underpins global trade and commerce, allowing countries, companies and institutions to trade, hedge and transfer risk.

Now a survey of over 200 FX trading institutions reveals that while 12% currently use algorithms, 38% plan to increase their use of algos in 2017.

JPMorgan believes 2017 is going to be “a watershed year for algo usage”.

In terms of currency mix, traders currently spend 70% of their time trading the major G10 currencies – including EUR, USD, GBP and JPY – and 26% in emerging markets.

This looks set to change in 2017 with 15% planning to increase their use of G10 currencies this year, with 32% planning to trade more emerging market* currencies as their liquidity continue to improve and they therefore become increasingly more attractive to trade.

So it’s no longer just about speed and a race to the bottom to be first in and out of the market – so called ‘bad algos’ beating everyone to the punchbowl.

The unifying theme of both the rise of the machines and the renewed interest in traditionally ‘less traded’ currencies is the search for liquidity in an increasingly fragmented and competitive market.

Algos can monitor and act across multiple venues, markets and currency pairs to flag opportunity or alert to risk.

Likewise, an uncertain macro-economic outlook plus improving liquidity makes trading in less-traded pairs much more attractive.

As the first signs of Donald Trump’s victory in U.S. presidential elections emerged the largest increase in currency pair activity was the U.S. dollar traded against the Mexican peso (USD/MXN), 63 times normal levels in the hour following the result.,

By way of comparison, spikes were also registered across the major currency pairs with input volumes ten times normal levels for EUR/USD for that hour, followed by USD/JPY and GBP/USD.

Turning to FX instrument type,40% of FX traders report that they plan to use more options in 2017, with a corresponding increase in cash, swaps and NDFs as hedging tools in an uncertain political and economic environment.

*On the whole at Chatsworth we’re not so keen on the term ‘emerging markets’ which is largely subjective and frequently inaccurate as many ‘emerged’ long ago.

Pragma launches SmartFix algorithm to improve FX trading performance against WM/R 4pm

Pragma Securities, a leading provider of high performance algorithmic trading tools, has launched a sophisticated algorithm designed to improve average execution performance against the daily 4pm WM/Reuters foreign exchange benchmark fixing.

Following recommendations from regulators in the wake of the FX rate-rigging scandal, in February 2015 the methodology underpinning the WM/R benchmark was changed, widening the calculation window from one minute for the most liquid currencies to five minutes.

In addition, banks have largely shifted their execution of customer fix orders from the spot desk to their electronic desks, where time-weighted average price (TWAP) execution algorithms are often used.

Pragma’s research highlights that together, these changes have created predictable patterns that can be leveraged to improve trading performance using only publicly available data. The full research findings are available via the Pragma Securities website.

Pragma’s new SmartFIX algorithm is built on the firm’s own research, which has identified predictable patterns of trading behaviour around the key FX benchmark.

The algorithm observes only publicly available information, and adjusts its trading rate in a systematic way based on those observations to achieve better execution on average for traders benchmarked to the Fix. In addition, these dynamic adjustments are layered on top of a proprietary trading schedule that achieves lower tracking error against the benchmark than a simple TWAP

David Mechner, CEO at Pragma Securities comments: “For traders that are constrained to match the fixing rate, our algorithm can reduce risk relative to a simple TWAP, and can also improve execution quality for a modest increase in risk. This makes it a good tool for banks servicing customer fix orders whether in a principal or agency manner.”

The new algorithm complements Pragma’s existing foreign exchange platform offering, Pragma360, which includes a suite of execution algorithms, transaction cost analysis (TCA), risk controls, and a next-generation algorithm monitoring system called Panorama. Pragma360 is provided as a broker-neutral trading solution to banks and asset managers.

Portfolio Items