how many of you here like taking
afternoon naps good god I’m in so much
trouble today I mean not about tell
o’clock all right the benefits of these
naps are well documented let me add a
little known benefit to this list assume
you are a u.s. investor who regularly
takes afternoon naps the chances of you
getting a cardiac arrest on the sixth
May of 2010 would have been
significantly lower than somebody who
doesn’t takes these naps sounds
interesting right here is why 6 May 2010
began like any other normal trading day
in the US markets market us up a bit
down a bit nothing much that we hadn’t
seen before so let’s say you took your
beauty nap at 2:30 p.m. and then woke up
at 3:30 p.m. right what happened to the
market the market didn’t do much right
so there’s nothing much to worry about
however if we had stayed awake during
this period this is what you would have
witnessed around 2:30 p.m. the markets
crashed dramatically the Dow Jones index
which is a proxy for the financial
health of the stock markets fell by a
massive 9.6 percentage as a result over
1 trillion dollars of investors wealth
was wiped out let’s put that number in
perspective right one trillion dollars
is twice the combined size of Singapore
and Hong Kong twice the combined size of
Singapore and Hong Kong economy was
wiped out within 15 minutes you should
have definitely taken that nap right
however if that sounded strange what
came next was even stranger but the next
15 minutes the market recovered and went
back to where it was earlier as if
nothing had happened right this crash is
referred to as a flash crash there’s
been six years since a crash happened
during the past six years we have
learned a lot not a lot about this crash
these learnings have in turn
fundamentally enhanced our understanding
of the financial markets
these learnings are by no means specific
to the US market
some of them apply even to the Indian
market where we are assembled today so
it is some of these insights that I
would like to share now let’s start with
a simple thought experiment right I’ll
give that you are sitting in a bar
that’s a good place to visualize
yourself in right now right so you
you’ve got a nice window conifer gonna
see for yourself after some time you see
your friend stepping out hey woody be
ten steps later oh let’s say fifteen
steps later of course it depends on the
amount of alcohol is consumed right
let’s assume that is consumed a lot of
alcohol
it’s going to be absolutely difficult to
predict where he would be ten steps
later right this part is often referred
to as a drunkard walk or in statistics
terms a random walk
many believe that stock prices are like
drunkard it’s very difficult to predict
their future prices so what is it that
we can predict let’s revisit the bar
right so now after stepping out the
drunkard our friend unchanged his dog
and they both start walking like before
it is very difficult to predict where
our friend would be but there is
something else that we can predict with
a lot of confidence right the distance
between the dog and the Drunken assuming
the drunkard did not treat the dog badly
the dog will always stay close to that
drunkard no matter where he won this off
stock market is full of dogs and
drunkards not literally of course right
what do I mean by this
now let’s say we have two stocks Google
and poodle poodle stands for a poor
man’s Google something that I made up
right so don’t look for it and Bloomberg
or something like that
so the key assumption here is that both
of them belong to the same sector have
similar business models and hence their
prices are driven by similar factors not
surprisingly they prices track each
other now let’s assume that over the
next two periods the price of poodle
falls while that of Google remains
stagnant the prisoners
what is likely to happen in the next
period before we make the call because
obviously scan use outlets for any
information relevant to poodle let’s say
we do not find any news or information
that would justify this dramatic crash
in price of poodle perhaps is just a
temporary aberration and eventually the
prices would converge what you’re
essentially betting here is that the
Google and poodle or the dog and
drunkard which eventually converge this
is what is referred to as space trading
as we are not looking at stocks in
isolation but in groups off took let’s
make it a bit more realistic right if
all of that I am doing is tracking these
two stocks I’m gonna spend a lot of time
praying that they would eventually
diverge because unless they diverge I
don’t have a trading opportunity right
however if I were to track three stocks
I can now work with three pairs if I
were to track hundred stocks symbol
algebra tells us that it can work with
close to 5,000 pairs my opportunity said
immediately explodes
however with great opportunities come
great pain it’s almost impossible for
the human brain to simultaneously
process information about 5,000 pairs
this is where algorithms step in and
algorithm is nothing but a series of
instructions to a computer right it’s
relatively straightforward to ask the
computer do the to the following read
prices of various stocks identify the
pairs that are moving away from each
other
and for the ones that have really moved
far away from each other place bets that
they will eventually converge this is
referred to as algorithmic trading a
practice where algorithms scan the
market identify opportunities and
execute trades all with no human
intervention right so it’s very
important to clarify that pace trading
is not the only algorithmic trading
strategy right now let’s make this
a bit more interesting now right let’s
add a bit more color what will happen if
I were to put these algorithms on drugs
or hallucinogens right what I get is
what is referred to as high-frequency
trading something that I’m going to talk
about now let me start with a very
simple illustration in US stocks are
primarily traded in New York and New
Jersey
however contracts that are tightly
linked to these stocks are traded in
Chicago these contracts could be simple
bets on whether the price is going to go
up or down given their tight linkages
it’s natural that the prices will move
together and hence they form the ideal
dog and drunkard pick an algorithm that
can continuously monitor the prices at
these two locations will be able to
exploit any temporary divergence in
prices by some estimates if a trading
firm can capture the monopoly of
exploiting these prices it stands to
make about twenty billion dollars in
annual profit that’s twenty billion
dollars in annual profit but how can a
firm capture this monopoly right
everything else remaining constant the
firm that is able to communicate the
fastest with these two exchanges will
have the upper hand it’s been surprising
that this has triggered a new kind of
arms race a race for higher and higher
speed training that happens at
exceptionally high speeds is referred to
as high frequency trading in 2012
Financial Times published a report by by
they said that hey cities account for 84
percentage of trading in u.s. stock
markets 84 percentage really investors
suggest pension funds mutual funds hedge
funds brokerage and other institutional
investors accounted for a mere 16
percentage it’s very tempting to
identify such prolific growth only with
Western markets like us the Indian
market where algorithmic trading was
introduced in the year 2008 presents a
perfect platform to validate this
hypothesis my co-authors and I have done
extensive research on this subject
in a project that was sponsored by New
York University stone school a National
Stock Exchange of India we found that as
early as 2010 algorithms accounted for
15% age of trading in big stocks
subsequent research shows that the share
had increased to 70 percentage by 2013
within five years this new breed of
traders had captured 70 percentage of
the market right the message is loud and
clear machines dominate trading in
markets this is the new normal
whether it is India or us machines
dominate trading in US markets let’s
next look at a dramatic illustration of
how this race for higher and higher
speeds has played out in the markets in
his bestseller flash boys the author
Michael Lewis introduces us to Dan
Spivey a trader based out of Chicago
like other traders he was very
frustrated with the poor transmission
speeds provided by the traditional
carriers how bad was it you may wonder
in 2007 it took 16 milliseconds for a
signal to do a two-way trip between
Chicago and New Jersey let’s put that
number in perspective right a blink of
an human eye takes 300 to 400
milliseconds thousand milliseconds is
one second right so by the time we blink
our eye a signal would have made twenty
round trips between Chicago and New
Jersey we might find this positively
baffling but the traders founded
woefully inadequate right so smiley
sensed a business opportunity if he
could find a faster way of communicating
data between these locations the speed
bandits would pay him a handsome reward
so he started a telecommunication firm
the name of the firm was spread networks
and his objective was to build a
straight a line as possible between
these two locations why because when spy
we started examining the cable roots of
the carriers we found that there are
lord of places where the roots were not
straight that makes sense right if you
have mountains or rivers the tables have
to go around them and not through them
however these twists and turns greatly
resolve in drop of transmission speeds
so spy we decided to build a straight a
line as possible
it meant literally blasting holes to the
mountains and digging tunnels under the
riverbed which he did finally Spivey had
a line through which he could transmit a
two-way signal within 13 milliseconds
right so if you are anchoring yourself
in the eye blink world we went from one
by twentieth of an eye blink to one by
twenty-fifth
of an eye blink mind boggling right so
how much did it cost by B to build this
build in life 300 million dollars just
so that hft traders could shave off
three milliseconds or one by hundreds of
an eye blink the arms race was truly on
despite this high stake race the beauty
about hatchapee was that they managed to
stay away from the public spotlight for
a very long time that is until the day
of the flash crash suddenly the acronym
hft started popping up everywhere and
the Google search for the keyword
hatchapee went through the roof that’s
what we do then you don’t know something
right we just go ahead and good so
everybody started looking what hey chip
D was but what actually did happen on
that day at 2:32 p.m. a mutual fund
group not a chip t remember send a huge
cell order to the market due to the
intense selling pressure the prices
started falling initially hfts viewed
this as the temporary aberration and
started buying but when the prices
continued to fall they sold whatever
they had bought earlier some of them
continue to aggressively south and some
decided to stop trading as a result
there is hardly any buyer in the market
and the market tanked so here is a
remarkable statistic between 245 13p
to 4527 p.m. a total of 14 seconds hft
is traded about 27,000 contracts or
roughly 2000 contracts per second or
thousand contracts per eye blink right
what was the role of the hfts in this
crash this is where a lot of ambiguity
is and that’s why we need proper
research to educate us a recent paper by
researchers from MIT at University of
Maryland
among other schools summarizes it nicely
they say while ketchup teas did not
cause the flash crash they contributed
to it due to their excessive training
this is not semantics right this is not
a trivial case of potatoes potatoes
tomatoes tomatoes it’s a sharp
difference between what they did and
what they were believed to have done now
so what is the learning what have you
learned in the past six years here is
what we have learned algorithms
specifically Hatcher T’s dominate
trading in most markets that’s the new
normal as soon as a flash crash
demonstrated this new normal
unfortunately is inherently unstable the
billion dollar dollar question or should
we say the trillion dollar question is
how do we improve the stability of these
markets researchers regulators
policymakers exchanges have thrown their
hats in the ring unfortunately we still
do not have a definite answer until we
find an answer we can only hope that
where every time HF please go on a
rampage like this we somehow managed to
doze off
this was 23rd April 2013 and around 1 7
at 107 on that day the markets witnessed
a micro crash losing about 200 billion
dollars in investors but what happened
on that day on that day The Associated
Press Twitter account was hacked so the
trading algorithms read the fake tweet
then realized that the tweet was fake
started selling and the market crashed
all of this happened within one minute
all of this happened before
Associated Press realized that his
account was hacked right
so in summary right in the olden days
what would that trade are done the
trader would have just picked up the
telephone and asked hey did they really
bomb the white house but now by the time
the trader picks up the phone unlocks
the screen dials the number and says hey
a gazillion orders would have gone to
the exchange The Shard is a crazy a fast
word out there thank you [Applause]