Download as pdf or txt
Download as pdf or txt
You are on page 1of 0

M a rc h 1 , 2 0 0 5

B u sin e ss In n o va tio n P o we re d B y Te c h n o lo g y
The 'HolyGrail':
Pre-Trade Analytics
By Daniel Safarik
I n today' s hyper-competi ti ve, cost-con-
sci ous tradi ng envi ronment, fund managers
and buy-si de traders are obli gated to watch
every penny, and to do so, many have turned
to computeri zed algori thms provi ded by bro-
kers at cut-rate commi ssi ons. Algori thms have
become a must-have for brokers seeki ng to
gai n new busi ness and retai n current cli ents -
TowerGroup esti mates that buy-si de adopti on
of algori thms wi ll tri ple between 2003 and
2006. But now that a plethora of algori thms i s
avai lable, the buy si de i s looki ng for more
quanti tati ve support than j ust post-trade
transacti on cost analysi s ( T CA) .
To meet that demand, the sell si de and the
T CA vendors are scrambli ng to provi de pre-
trade analyti cs, whi ch, through the analysi s of
hi stori cal and current pri ce and volume data,
attempt to help cli ents determi ne where to
send orders and when; whether to use algo-
ri thms or manually trade an order; and what
the opportuni ty cost of not acti ng on i nforma-
ti on mi ght be. But brokers have been strug-
gli ng to produce pre-trade analyti cs that are
meani ngful for cli ents and have encountered
trouble i n di stri buti ng these analyti cs to cli ent
desktops, both for cultural and technologi cal
reasons.
At the moment, most pre-trade analyti cs
products cannot accurately predi ct the pri ce
of a stock for a gi ven day, whi ch i s what
traders ulti mately want, accordi ng to Li nda
Gi ordano, product speci ali st, Quanti tati ve
Servi ces Group ( QSG) , a T CA vendor. Most
compare the spread between bi d and ask
pri ces, reference that agai nst the volati li ty of a
gi ven stock and attempt to create a range of
potenti al outcomes.
Usually, the analyti cs predi ct the expected
cost i n tradi ng fees that cli ents would pay i f
they executed a gi ven tradi ng strategy at a
gi ven poi nt i n ti me, says Gavi n Li ttle-Gi ll,
author of a TowerGroup report on algori thmi c
tradi ng.
"Pre-trade analyti cs i s a Holy Grai l that has
been very elusi ve, " QSG's Gi ordano says. "I t i s
tough to predi ct i n a way that i s relevant or
practi cal, and i t i s almost i mpossi ble to do i n
the way the market expects to do i t."
To Build or Not to Build
Most buy-si de cli ents get thei r analyti cs from
brokers as part of an overall servi ce package.
T he benefi t to thi s i s that, havi ng created the
algori thms, brokers know them backwards and
forwards. However, there i s a percepti on that
brokers offer bi ased analysi s that favors thei r
own algori thms. Addi ti onally, buy-si de fi rms
are concerned wi th di sclosi ng too much i nfor-
mati on about thei r propri etary tradi ng strate-
gi es. As a result, buy-si de fi rms often choose
to use thi rd-party T CA vendors or bui ld pre-
trade analyti c capabi li ti es i n-house.
Most T CA fi rms, however, provi de only
post-trade analysi s, though QSG i s developi ng
a pre-trade offeri ng, Gi ordano notes. T hese
provi ders can be valuable i ndependent tech-
nology sources, but thei r soluti ons are li mi ted
by the i nformati on that brokers and the buy
si de are wi lli ng to share.
So, some of the larger, more technologi cally
savvy buy-si de fi rms conduct thei r own analy-
si s, whi ch requi res a ti ghtly i ntegrated
research and tradi ng team and substanti al
amounts of real-ti me data. For example,
Barclays Global I nvestors ( BGI ) conducts all
of i ts analysi s i n-house, updati ng stock-speci f-
i c models dai ly wi th real-ti me data feeds and
trade hi stori es from i ts i n-house order man-
agement system ( OMS) and i ts FlexTrade
executi on system, accordi ng to Ananth
Madhavan, global head of tradi ng research.
"T he concept of more sophi sti cated pre-
trade analyti cs i s i mportant, " says Mi chael
Sobel, head of U.S. equi ty tradi ng, BGI . "But
the essence of brokers wi th T CA i s that they
do have advance knowledge of what you are
doi ng - we try to avoi d that."
To li mi t di sclosure of thei r tradi ng strategi es
when usi ng pre-trade tools provi ded by bro-
kers, some buy-si de fi rms run the software i n-
house rather than send guarded data to the
brokers. "Any [analyti c system] that we use i s
used on-si te, because i f we shot a li st of
[i ntended trades] to a broker and had them
analyze the li st, we'd ri sk exposi ng the li st. I
feel we can get sni ffed out i n the market, " says
John Wheeler, di rector of U.S. equi ty tradi ng,
Ameri can Century I nvestments. Wheeler adds
that, si nce i t has bui lt i ts own tradi ng systems
wi th di rect data feeds, Ameri can Century can
perform i ts analysi s on raw market data,
rather than rely on a broker- or vendor-sup-
pli ed tradi ng system to supply and i nterpret
that data.
Benchmark Ambiguity
I t's not only the suspi ci on of vested i nterests
that has made marketi ng analyti cs a chal-
lenge for brokers; there i s also no prevalent
standard benchmark for evaluati ng algo-
ri thms agai nst each other - and some say
there shouldn't be.
T he most prevalent tradi ng benchmark i n
use today i s volume wei ghted average pri ce
( VWAP) , whi ch i s calculated by addi ng the
dollars traded for every transacti on i n terms
of pri ce and multi plyi ng that by shares trad-
ed, then di vi di ng that by the total shares
traded for the day. VWAP i s popular because
i t's relati vely easy to measure and provi des
comparati ve results, but i t i sn't as useful for
evaluati ng strategi es that are tryi ng to do
somethi ng other than follow the market mi d-
poi nt; such strategi es often acti vely trade
si ngle stocks, usi ng algori thms that can be
altered on the fly.
"VWAP i s really the only agreed-upon
measure out there, " TowerGroup's Li ttle-Gi ll
says. "T he problem wi th applyi ng that to
algori thms i s that i t doesn't take i nto consi d-
erati on anythi ng other than i ndexi ng. A
thousand shares of I BM traded four ti mes i n
a one-week peri od wi ll get a di fferent execu-
ti on quali ty each ti me. People talk about
compari ng algori thms across di fferent bro-
ker-dealers, but i t i s really tough unless you
have si gni fi cant quanti ti es traded, and i t i s as
much an art as i t i s a sci ence."
VWAP i s often i ncorrectly appli ed to
measure the performance of algori thms that
have di fferent tradi ng goals, accordi ng to
Bri an Fagen, managi ng di rector at Morgan
Stanley, who adds that algori thms should be
evaluated as part of a fi rm's overall strategy
to achi eve a goal, not as separate enti ti es.
T he desi re for a benchmark speci fi cally for
algori thms "calls i nto questi on those algo-
ri thms that do not have a goal, " he says.
Brokers have begun to focus on other
standards, i ncludi ng market-on-close ( MOC)
and arri val pri ce. MOC measures the last
pri ce obtai ned by a trader at the end of the
day agai nst the last pri ce reported by the
exchange. Arri val pri ce i s the mi dpoi nt of the
bi d-offer spread at order-recei pt ti me, and i t
also notes the speed of the executi on.
Rob Flatley, managi ng di rector of Bank of
Ameri ca's Electroni c Tradi ng Servi ces ( ET S)
uni t, says the market i s movi ng toward these
new standards. "T he quant-est shops on Wall
Street have already moved to arri val pri ce, "
he says.
T here are algori thms geared to all three
benchmarks, but a "pure" benchmark of all
the algori thms i n the market i s unli kely to be
created, si nce brokers aren't keen to part
wi th i nformati on about thei r tradi ng partners
or strategi es. "We do not collect the i solated
data that would be needed to rate algori thms
agai nst each other, between brokers, " says
Mari e K onstance, di rector of sales and prod-
uct management at Plexus Group, a T CA
vendor. "Brokerage fi rms do not provi de
algori thmi c tradi ng data on i ts own; you only
can analyze how well they are doi ng overall."
"I t would be hard to mai ntai n [a generi c
algori thm benchmark] and i t would have
very li ttle relevance, " M organ Stanley' s
Fagen says. "I f my algori thm i s benchmarked
to the market close, does i t make sense to
measure that algori thm up agai nst some
other benchmark?"
Integration Obstacles
A lack of technologi cal i ntegrati on wi th buy-
si de OMSs also has restrai ned the use of pre-
trade analyti cs. Although broker-sponsored
tradi ng systems have algori thms and analyt-
i cs bui lt i n, very few vendor and i n-house
OMSs support the real-ti me ti ck data that
allows for i nformed, on-the-spot deci si ons.
"We've all had a tough ti me [li nki ng deci -
si on tools to] our executi on capabi li ti es,
because we are dependent on OMSs [that]
mostly weren't bui lt at the ti me algori thms
were i nvented, " Bank of Ameri ca's Flatley
says. But that i s begi nni ng to change.
BofA now i s i ntegrati ng deci si on support
tools wi th buy-si de order management sys-
tems, rather than deli veri ng the tools i n a
stand-alone screen off to the si de, a j ob that
i s about 50 percent fi ni shed, Flatley relates.
He says he expects to i ntegrate BofA's ana-
lyti cs appli cati on wi th seven to 10 vendor-
bui lt OMSs by the end of the year. But that's
only the begi nni ng of the i ntegrati on effort:
Flatley esti mates that 80 of the top 500 U.S.
mutual funds run propri etary OMSs that
would requi re a broker-dealer to make i ndi -
vi dual confi gurati ons for each cli ent.
Even as the i ndustry struggles to deli ver
effecti ve pre-trade analyti cs, the demand for
accurate predi cti ve i nformati on i sn't goi ng
away. Rather, i n the current hi gh-frequency
tradi ng envi ronment, the demand for pre-
trade analyti cs wi ll only grow. T hough ti ghter
technologi cal i ntegrati on i s bound to come i n
ti me, i f pre-trade analysi s i s to fully come to
frui ti on, fi rms may also need to fi nd a way to
share more i nformati on wi thout putti ng thei r
tradi ng strategi es at ri sk.
What They
Have to Say About
Pre-Trade Analytics
and TCA
"A bi g area of growth for us i s cli ents com-
i ng to us aski ng how to best use the algo-
ri thms we have, " says Carl Carri e, vi ce presi -
dent of new product development at
JPMorgan Securi ti es. "Once you are usi ng an
algori thm, under what context do you use i t?
T hat whole si de i s becomi ng very useful. We
offer pre-trade analyti cal products over the
Web and embedded i nto executi on systems."
"A lot of the broker-sponsored algori thmi c
tradi ng systems attempt to measure or proj -
ect the trade costs as well, but they are pret-
ty rudi mentary, " says John Wheeler, di rector
of U.S. equi ty tradi ng, Ameri can Century
I nvestments.
"Pre-trade analyti cs attempt to predi ct
how close you could come to some ki nd of
stri ke pri ce, li ke the pri or day's close or
today's open, " says Eugene Noser, presi dent
of Abel/Noser. "None of them are very good.
Most pre-trade analyti cs overesti mate the
cost of a routi ne or beni gn trade and greatly
underesti mate the cost of di ffi cult trades."
"I f you thi nk of basi c ti me-sli ci ng, the abi l-
i ty to map that acti vi ty and mai ntai n that
benchmark i s what separates people, " says
Jana Hale, global head of algori thmi c tradi ng
at Goldman Sachs. "You need to do a li ttle
[analysi s] i n-house, a li ttle wi th the thi rd-
party T CA and a li ttle wi th the sell si de, and
wi thi n that tri angle of resources you come
up wi th best executi on."
"I n a stock that i s not so li qui d, i f you trade
a large posi ti on over the course of the day, i f
you are measured agai nst VWAP, you
become the VWAP, " says Li nda Gi ordano,
product speci ali st at T CA fi rm Quanti tati ve
Servi ces Group ( QSG) . "T here i s behavi or
that goes on i n the mi crostructure of a trade
that makes VWAP a very poor, yet prevalent
strategy. I t's become li ke a reflex."
"I t's as cri ti cal to have a deci si on support
system i n an algori thmi c sui te as i t i s to have
executi on capabi li ty, " says Rob Flatley, man-
agi ng di rector of Bank of Ameri ca' s
Electroni c T radi ng Servi ces ( ET S) uni t.
"From the pre-trade perspecti ve, we thi nk
arri val pri ce or market-on-close ( MOC) are
the two standards that people have already
started to focus on versus VWAP."
Copyright 2005 by CMP Media LLC, 600 Community Drive, Manhasset, NY 11030. Reprinted from WALLSTREET & TECHNOLOGY with permission. 5955

You might also like