Jun 26, 2012
Squeezing out every last drop
Entrusting the
execution of liquid trades to third-party algos to free up buy-side traders’
time could result in a meaningful reduction of returns for institutional
investors, according to Ethan Levinson, president of agency brokerage SJ
Levinson & Sons.
The New York-based
agency broker has attempted to help the buy-side add alpha to its simpler, more
liquid trades – commonly referred to as ‘noise’ – through the latest version of
its Strategic & Tactical Analytic Research & Trading (START) automated
US cash equities execution platform.
Levinson says the
concept behind START was to improve the automated execution of trades buy-siders might have marked up a couple of pennies and released to a
third-party algorithm in the hopes of freeing up space on their pad and
declaring success if the stock traded up a dime. In The TRADE’s 2012
Algorithmic Trading Survey, increased trader productivity was identified as one
of the top three reasons for using algos, behind reduction of market impact and
ease of use.
However, missed opportunities to capture those unrealised
pennies over time could eventually add up to basis points and then to real
returns – or lack of returns – for money managers, says Levinson, explaining
the rationale behind START.
"Essentially the platform is a co-trader,"
he says. "If a trader receives a basket of orders, how many of
those can he really focus on and trade well? Ten? The platform's engine can take
up to hundreds of orders based on rules that are set up by the client in
conjunction with our research team. It then takes a percentage of those orders
and starts trading them automatically, freeing up the trader to work on the
more difficult trades."
Taking this alternative approach to the execution of liquid
orders can reveal differences between orders that look similar but may need to
be executed in different ways.
"If you get two orders for the same security – for instance, IBM – and put them into an algorithmic strategy like volume-weighted average price (VWAP),
there won't be a discernible difference,” he adds. “With START, you can wind up
with two unique strategies dependent on the characteristics of the order and
how we profile the order."
The START platform was initially rolled out in mid-2010 and
consists of a series of intelligent rules engines which sit on a FIX messaging
gateway and evaluate every submitted trade using a number of variables to
create a custom execution strategy, which is then executed via child orders
sent to the appropriate market using direct market access (DMA) connections. Since
portfolio managers route orders directly to the START engines, which send them
to the market via DMA connections, there is no chance of information leakage
during the trade.
The latest version cuts down the implementation time from
approximately six months to six weeks thanks to a new framework which eliminates much of the “heavy lifting” involved in earlier
deployments, says Levinson.
Always learning
At the heart of START are a number of execution rules
engines that continuously learn and adapt, depending on the amount of order flow
that runs through them.
"It's not an issue of sitting down and setting up the
platform once," says Levinson. "You could do that, but the engine is
constantly learning and getting better depending on the flow that goes through
it."
Every night, in an automated fashion, the platform sends a number of performance
files to SJ Levinson's research team to be reviewed, updated and sent back to
the platform, which resides in the client’s IT infrastructure, in time for the
next day's trading session.
SJ Levinson staff help client firms configure the engine
from its default settings to the clients' desired configuration by using data
pulled from their trade-cost analysis (TCA) platforms, such as SJ Levinson's
Trade Analysis Platform (TAP) trade to set up the rules, for example sending
the orders that are less than 10% of the average daily volume, to START
directly.
"Client can use TAP to focus on trading flow and to
configure START – they're complementary, but one is not plugged into
another," says Levinson.
After that, there is nothing left for the buy-side trader to
do with the START-directed trades.
"The order goes into the engine and then out those DMA
connections into the marketplace," says Levinson. "Nobody sees the
order. They're sent out in small DMA slices.”
Depending on the nature of the trade, whether it is in a
highly liquid or illiquid issue, the strategy could range from 20 minutes to
four days, he adds.
Buy-side clients have a choice of implementing the platform
via an API to integrate it with existing execution platforms or as a separate
screen displaying various configuration knobs and switches.
The platform is also capable of cross-border trades with
other equities markets as long as there are the necessary DMA connections to
the market. However, the platform does not currently handle currency conversion
for cross-border trades. Foreign exchange will be the next instrument class
supported by the system, says Levinson.
Rob Daly