Coaching users in algo techniques
| — Stavros Siokos
Stavros Siokos, head of European
alternative execution, sales, Citigroup tells FTMandate about
educating the buy-side in the best use of algorithmic trading
FT Mandate: With algorithmic
trading seemingly besieging stock markets on both sides of the Atlantic
over the past year, do you think the buy-side users still need
(ongoing) education about algorithmic trading techniques in order to
make the most appropriate use of algorithms and reduce the fear among
some over the so-called ‘black box’ and even ‘white box’ applications?
Stavros Siokos: Absolutely. We are working on new ways to help the
buy-side to make the most of our algorithmic trading product. Recently,
we have been using our internal traders more as coaches – these guys
use our algorithms day-in, day-out and so are best placed to transfer
some of their knowledge about how they work to our clients.
Our approach is very much from a
white, or ‘glass’ box point of view – ‘black’ box strategies should not
be used by the buy-side, as by definition if their behaviour is hidden
from the user it is not clear what they are going to do.
We’re still learning as well – much of
the usage of algorithms and strategies differ between buy- and
sell-sides, and we work constantly with our clients to develop our
algorithms to suit their needs better.
FTM: Clearly there are different tastes and reasons for using
algorithms as an execution and trading tool. In terms of the perception
(and reality) of algorithmic trading and DMA (direct market access),
how does it vary from institution to institution on the buy-side? Has
this changed over the last six-12 months and in what ways has Citigroup
been responding? Are algos being used for ‘high’ as well as ‘low touch’
SS: We have seen in recent
industry surveys that there is a phenomenal appetite for algorithms,
but perhaps only a few per cent of overall European flow is traded via
DMA and direct algorithm access in Europe. While this is growing,
buy-side technology is a limiting factor in this. This tends to be the
main factor in differentiating between algorithm take up among the
buy-side: while some institutions are at the forefront of the
technology race – with advanced order and execution management systems,
others, often owing to budgetary constraints, are not yet FIX enabled.
The buy-side has moved forward over the last year, but the pace has
perhaps been slower than anticipated. We are seeing ‘high’ touch
business increase as the comfort level goes up – our internal trades
have been putting large- or mid-cap orders on our algorithms for years
and we see no reason why the buy-side should be different.
FTM: With algorithmic trading accounting for around 50
per cent of trading in large caps in the US, possibly 10 per cent to 20
per cent in Europe, and Asia coming from a lower base, what can we
expect from these different regions over the next 12 months in terms of
growth and adoption of algos? What potential does Asia-Pacific have and
which countries will be the first movers in take-up ?
SS: There’s little doubt that
we will see algorithmic trading gain market share globally over the
coming months, as electronic markets become more liquid and buy-side
comfort level increases. This growth will be further boosted, as more
and more assets flow into hedge funds, which tend to be the primary
This growth will be slower in the US
than the rest of the world, while in Europe as buy-side technology
improves we expect growth to take off over the next couple of years.
Furthermore, increased scrutiny, regulation and accountability will
tend to push the buy-side towards more algorithm usage owing to the
transparency of their execution performance and cost effectiveness.
FTM: With so much talk about multi-asset trading and
gaining arbitrage opportunities from trading across asset classes, how
appropriate are algorithms for that and for trading mid- and small-cap
equities, emerging market equities, FX and commodities? Is more
development needed in terms of IT infrastructure and capacity of
systems (connectivity, FIX) to allow this to really bloom going forward?
SS: In the equities world,
algorithms can trade just about any stock, provided the right algorithm
is chosen, and the model behind it is well written and tailored to the
market. In respect of true ‘arbitrage’ type strategies, the sort of
funds or prop desks which trade this will typically write their own
algorithms and will have little demand for what brokers would offer.
Outside equities, the demand for ‘execution’ type algorithms is likely
to be limited in the foreseeable future, among other things many
instruments are not even traded electronically and they tend to be net
markets, so commission is not an issue.
FTM: Despite a plethora of algorithms and solutions
being touted by various providers – from agency/brokers and sellside
firms – how important is it for service providers and sellside firms to
differentiate themselves in terms of product and service offering,
connectivity and innovation? How does Citigroup differentiate itself
and what is the current strategy across the globe?
SS: It’s vitally important –
getting the right outcome from using algorithms is as much about
service, support and consultancy as the algorithms themselves. Globally
we have both algo development and support teams in each major hub: so
that solutions are tailored for the market in which they are going to
FTM: In the US, with RegNMS coming up this July (MiFID
in Europe too) and the provision for ‘best execution’ (trading on an
automated venues, likely fragmentation of execution), what impact will
this have on institutions trading algorithmically? How important is it
likely to become with RegNMS mandating use of automated trading venues
and what should buy-side institutions be considering now?
SS: We expect that all these
regulations will increase fragmentation of liquidity, both in the US
and Europe. This can only encourage the uptake of algorithms among the
buy-side – there is no way that a human being can simultaneously
monitor multiple liquidity pools. We have no doubt that the smart order
routing algorithms that are prevalent in the States will become a
must-have in Europe for the buy- and sell-sides.
FTM: With such a focus on performance, trading costs
really matter to investors and can make a significant difference in a
portfolio’s overall return. What do you consider the key factors in a
decision to deploy one algorithmic trading system over another – ease
of use, connectivity (FIX), liquidity, best execution and transparency
or a combination?
SS: Ease of use, flexibility
and transparency are key differentiators for us. In some ways, ease of
use and flexibility are pulling in different directions – it’s hard to
have a product that can achieve a wide variety of execution objectives
without a large number of parameters. The way we get around this is by
working hard with our clients to ‘can’ popular parameter sets so that
they can be achieved with just a single click, or even working on
custom strategies. Transparency of performance is important to us –
that’s why we use BECS post trade, which has a large universe and rich
peer comparison to show which brokers are doing a good job for their
FTM: Education, education, education – that was a key
theme from the recent TradeTech event in Paris in relation to the needs
of the buy side. How does Citigroup enable its clients to recognise the
value the organisation brings to them as ‘execution’ consultants or
specialists who dispense practical, ‘real world’ trading solutions in
Europe, North America and Asia 24/7? What is your role as execution
consultants to clients and various institutions?
SS: Algorithmic trading forms
part of the Alternative Execution group here at Citigroup – including
Program Trading, DMA and portfolio strategies. Our sales people have an
outstanding feel for these products, and so we are able to help them
choose the right mix of these and our cash and derivatives businesses
to maximise their alpha capture.
Our Global Portfolio Trading
Strategies (GPTS) team has been consistently ranked number one in a
variety of surveys for their portfolio analysis service – their role is
now broadening to include all aspects of the investing and execution
We also provide a bi-monthly marketing
piece called ‘Inside the Box’ with performance statistics, news and
tips to get the most out of our product.
FTM: ‘Best execution’ is a term that has many different
definitions. Some say it means delivering the execution that best suits
the client’s needs. How do Citigroup’s algorithmic trading solutions
help in that regard? Clearly algos are not appropriate for every trade,
but can help deal with the ‘noise’ on trading floors and simpler trades
– freeing up traders to intellectualise over the more complex
transactions. Can they help generate alpha too?
SS: Best execution is a term
that gets thrown around and used for many different reasons. It’s
possible to satisfy the regulators’ definitions, without providing the
best execution, which is what we are always striving to achieve for our
clients. Algorithms can be used to automate simple orders – but also to
take advantage of fleeting opportunities that may exist for
milliseconds. It’s unlikely that any broker would give out a model that
‘generated’ alpha for a few basis points – but there’s no doubt that
algorithms used well preserve alpha compared to a pure manual process.
We’ve seen this internally over many years – our traders have much more
time to deal with their difficult trades or capital commitment.
FTM: Despite a plethora algorithmic flavours on the market, in your
experience would you say clients are finding that there is a limit in
terms of the number and ways in by which clients execute their
algorithmic business? Is that dictated by trading style and/or
SS: Typically the standard
strategies – VWAP, TWAP tend to dominate our flow – albeit with
parameter overlays. There is a near infinite way to put these
parameters together, but most clients will find a set that work for
them and integrate well with their investing process. The biggest
barrier to new strategies tends to be the lead time in getting OMS’s
updated to include the up-to-date offering.
FTM: Where does Citigroup envisage developments in
algorithmic trading strategies and execution to be heading over the
remainder of 2006? And, what can we expect in non-equity algorithms –
FX and futures, algorithms of algorithms, and listening or polling
algorithms? Will customisation and even analogies to biological
science/nature play a role in developing killer algo strategies? With
that in mind, are the days of the human trader numbered or is that
SS: Algorithm use will continue
to grow, and we’ll probably see more demand from sophisticated clients
for tailored algorithms. With regards to emerging markets and less
liquid stocks – we’re already there, and have been trading mid-caps on
algorithms for a number of years. Multi-asset class algorithms, as I
mentioned earlier, are unlikely to take off in the same way as their
equity counterparts. Human traders will always play a role – it’s just
that the skills demanded for that role evolves. The big bang of 1986
changed the way the stock market operated in London for ever, and
successful traders adapted to that change. Although it’s a gradual
evolution, we’re seeing something similar here with algorithms, DMA and
FT Mandate: Aside from the view of some institutions that not all algos
do what they claim, transaction cost analysis (TCA) is key pre- and
post-trade, but should such measures be agency broker TCA offerings or
more independent solutions? Then there are the benchmarks - VWAP, TWAP,
INLINE. What makes most sense and why when analysing trading strategies?
SS: In BECS, we are able to
offer our clients a solution that offers the best of both worlds – we
have a world-class impact model, developed in collaboration with
academia: and an arms-length, independently audited post-trade
solution. The choice of benchmark depends on what is trying to be
achieved – typically we provide our clients with performance against a
comprehensive list of benchmarks, so that the algorithms can be
assessed from a variety of different angles – VWAP strategies should
obviously be assessed primarily against VWAP, but Arrival Price (the
mid-point when the order is started) should also be used to assess
information leakage and impact of the strategy.
In association with CitiGroup.