Angling with algos in unfamiliar waters
Good fisherfolk know the right
tackle to land the big catch – not every lure works in every pond. So if you’re
dipping your line in emerging market pools, what electronic bait is available
and where will it work best? Elliott Holley wades in.
Algorithms
exist to help buy- and sell-side market participants automate workflow, reduce
market impact and improve performance relative to objective benchmarks. In
developed markets such as Europe, some observers estimate they account for up
to 80% of order flow. But buy-side traders looking to use algos in the world’s
emerging markets find a more varied landscape, full of challenges and
opportunities.
Where to trade
Put simply, some emerging markets are too shallow. A
market must be sufficiently liquid and sufficiently electronic to justify algo
trading. Some markets, such as Brazil, have made consider- able progress in
attracting international investors through a combination of good governance,
investment in trading infrastructure and a strong underlying economy. Others,
such as the ‘four little dragons’ – Thailand,
Indonesia, Vietnam and Malaysia – are at an earlier stage of capital markets
development.
Estimates provided by financial research firm
Greenwich Associates suggest algos account for around 17% of buy-side equity
flows in pan-Asia, with the leading markets in Japan, Singapore, Hong Kong and
Australia experiencing algo flows closer to 20-25% and upwards. Algo trading
and co-location accounted for 25% of derivatives volume and around 30% of
equities volume on the National Stock Exchange (NSE) and Bombay Stock Exchange
(BSE) last year.
“Right now, there are algos participating in
most of the markets across Asia,” says Ofir Gefen, head of research and
algorithm consulting, Asia Pacific at agency broker and technology provider
ITG. “They are not so prominent in the more frontier markets like Vietnam or
Sri Lanka but they are more active in the borderline developed and emerging
markets, such as India, Korea, Taiwan, Thailand, Malaysia and Indonesia.”
Notably, the past 18 months has also witnessed a
steady flow of technology offerings into Brazil, including connectivity
solutions, algorithms, direct market access services, data centres and co-location
facilities. JP Morgan introduced a suite of trading algorithms in Brazil in
November, while SunGard and ITG have continuously added their own solutions to
the mix. Meanwhile beyond Brazil, the Mercado Integrado Latino Americano – a
regional integration project between the exchanges of Chile, Colombia and Peru
– offers international investors the prospect of untapped alpha that is
increasingly accessible to algorithmic strategies, particularly now that
S&P has released its MILA 40 Index of the most liquid companies in the
region.
“The top growth markets for the use of algorithms
globally are Brazil, Indonesia, Malaysia and the Philippines,” says Frank
Troise, global head of electronic client solutions at J.P. Morgan.
In the northern hemi- sphere, Russia has attracted
the attention of high-frequency traders using arbitrage strategies to profit
from price differences between the country’s merged MICEX- RTS exchange and the
London Stock Exchange (LSE). But while Russia has liquid stocks such as Gazprom
(the most traded stock on the LSE), the fact that the domestic bourse has to compete
with the LSE’s international order book – which lists the larger Russian
companies – proves that a range of factors can come into play.
Restrictions on currency may apply. T+0
settlement may complicate post-trade. Market participants may be using American
depository receipts instead of local
products. In the grand context, algos can
provide benefits but there are a whole host of other factors for investors to
take into consideration.
Which algorithm?
Once a given country or emerging market region
has been declared fit for algos, investment institutions must next decide which
tools to use.
According to David Jenkins, business solutions
manager, Asia Pacific at technology provider Fidessa, most of the demand in
Asia’s growth markets is for benchmark algos that can help reduce market impact
by helping traders navigate relatively illiquid markets and wide spreads. Algos
like TWAP, VWAP and percentage of volume are very popular, especially among the
local investment community, while international investors from the US and
Europe tend to use arrival price-based strategies.
However, even VWAP can be difficult to apply to
some markets, such as Thailand and Malaysia, because while the volume curve
doesn’t have huge variation in Europe, making it easy to develop benchmark
algos, in Asia it varies considerably.
“The morning trading one day could differ dramatically
from the next,” says Jenkins. “People tend to trade passively, and crossing the
spread is more risky in Asia. How do you deal with these conditions? Positioning
orders ahead of time allows you to have orders already waiting at the top of
the book.”
ITG’s Gefen believes that VWAP still accounts for
between 60-70% of algorithmic flows in Asia, down from approximately 80% three
or four years ago. Meanwhile, the lack of trading venue competition across many
emerging markets reduces the usefulness of opportunistic algos that work by
seeking out price improvement on different venues.
The lack of customised algos that are properly
adapted to specific emerging markets continues to be cited by many buy-side
firms as a problem.
“Clients get frustrated at the offerings in
places like Hong Kong and Brazil,” says Paul Daley, head of product development
at SunGard’s Fox River Execution business unit. “They will complain that the
algos they are receiving were built for the US. It takes time to build new
algos specifically for the emerging markets; some providers prioritise time to
market at the expense of quality. But clients are telling us that there’s too
much white labelling of essentially the same algos.”
Brokers must carefully weigh up the cost of
developing customised algorithms for every emerging market.
Accessing local market data feeds is essential
to building an algorithm that will be successful in a new market, according to
ITG’s Gefen, as algorithms need to feed on both historical data which show how
listed stocks have traded in the past, and incoming real-time data showing
immediate market trends and movements which might affect a trade. But some
observers suggest while market data is relatively cheap and commoditised in
London, in many markets across Asia financial institutions can only guarantee
high-quality market data by paying for co-location across the region – an
expensive proposition which may be too much for some firms.
The value of each market has to be balanced
against the cost of building co-location facilities and developing algorithms
for each specific market. For countries like Thailand and the Philippines, the
case is harder to make.
Adapting algorithms
Nevertheless, simply copying an algorithm
developed for the US and dropping it into a market in Asia is unlikely to
produce a satisfactory result for investors. Wider spreads in many emerging
markets would unhinge many vanilla developed market algos, leading to poor
execution choices.
“Algos from the US will cross the spread a lot,
which doesn’t matter in Europe, as you may only be paying one basis point,”
notes one observer. “But in Asia the spreads are a lot wider so it does matter.
It takes a lot more work to make an algo work in Asia.”
ITG’s Gefen concurs. “You can’t just take a US
algo and put it in Asia,” he says. “It will work, but the performance compared
to an algorithm that was calibrated to work in Asian markets would just be
horrendous. The market micro-structure is different to that found in the US or
Europe, so the algorithms have to operate differently. Hong Kong has no closing
auction, while in Japan, there are special uptick rules for short-selling – all
of which could cause problem for a vanilla US algorithm.”
Instead, algo providers study spreads,
volatility and volume profiles in each emerging market, as well as the exchange
policies and local regulations, to develop suitable algos tuned to the local
book depth and trading rules.
Regulatory factors must be taken into
consideration – however electronic the market. Gefen explains that Korea is a
highly electronic market, but the regulator frowns on firms which send high
numbers of correction messages to the market and will sometimes force offending
firms to explain themselves. To adapt and succeed in the country, algorithms
are configured not to go above a certain threshold in terms of numbers of
messages, so as to not alert the regulator.
In India, the Securities and Exchange Board of India
(SEBI) has expressed concern about the drive for ever-lower latency. SEBI
recently revised its guidelines for algo trading to mandate exchanges to have
systems in place to cope with algo trading, including blocking order flooding
and creating financial disincentives to discourage high order-to-trade ratios
from brokers. Among other measures, the exchanges must install systems to
identify ‘dysfunctional algos’ that lead to a ‘loop or a runaway situation’ and
shut down brokers’ terminals if necessary – so brokers need to be careful about
which algo strategies they allow clients to use in the country. The regulator
also maintains the right to look at any algorithm in detail.
Beyond the rules, there are also drastically different
market realities to adapt to. ITG’s Gefen points out that maintaining place in
the order queue is much more important in Asia’s emerging markets than in the
US – providing another potential pitfall to which an uninitiated algorithm
could fall victim.
“At any particular price on the order book –
especially at the top of the book – a queue forms because it’s based on price
time priority. So if you put your order first, it stands nearer the beginning of
the queue,” says Gefen. “In Asia, the queue is longer and slower moving. A typical
US algo will cancel orders and correct them many times a minute or even per
second. In Asia you have to be a lot more careful about doing that, because on
some of those corrections you could lose your place in the queue.”
Algorithms in Asia will also put more legs out
into the market, to position themselves in the queue. If a client corrects an
order upwards, the broker should not necessarily update immediately across that
position in the market, because that may cause the firm to lose its queue position.
Putting multiple legs in the market lets a buy-side institution maintain
different positions in the queue, correcting orders by using the last place in
the queue – without sacrificing the parts of the order that are already closer
to the front of the queue.
In addition, Jenkins at Fidessa explains that
retail flow uses synthetic orders in markets where advanced order types such as
iceberg orders are not yet supported – meaning algorithms need to be ready to
take account of local synthetic orders, especially since retail activity is a
major component of many of those markets.
Whatever the market and whichever the
algorithmic strategy selected, the world’s growth markets are increasingly
offering the chance to obtain untapped alpha. As electronic trading expands
across the globe, exchanges improve connectivity and brokers roll out new algorithms,
traders have more tools to access more liquidity than ever before, according to
Philippe Carré, global head of connectivity at financial technology provider
SunGard.
“From Santiago to Istanbul to Singapore, the
market structure is changing,” he says. “Technology is simplifying and
automating the trading process. Once you have the tools to trade a market, it’s
a virtuous circle. In markets like Turkey, more people are trading, volumes are
increasing – the opportunities are growing all the time.”