Darren Jer, CEO of MarketFactory, talks to Galen Stops about flash crashes, the new latency
arms race and how technology will enable the FX market to keep growing in
Galen Stops: What’s going to be the main focus for
MarketFactory as a company in 2017?
Darren Jer: Well let me just start by saying that FX is the
biggest market that not everyone knows about. In the equities
market last year, $114 trillion was traded across all exchanges;
in FX, that figure is $1.4 quadrillion. In FX we talk in average
daily volume (ADV) numbers all the time so we’re just used to
the size of the market, $5.1 trillion per day, but the general
public and traders in other asset classes don’t know the
degree of notional liquidity. So at MarketFactory our focus is
looking at how we can create solutions that encompass the
entire industry and make the market bigger for everyone, how
do we make the industry more transparent so that it invites
more people in and makes it easier for traders and investors
from other asset classes to come in and get involved?
GS: With the latest Bank for International Settlements (BIS)
figures showing that the FX market decreased in size for the
first time in 15 years, doesn’t this suggest that fewer people
are getting into FX?
DJ: The idea that the FX market has shrunk is actually
erroneous. It hasn’t shrunk because at the end of the day the
FX market is driven by direct foreign investment and global
trade and by speculators in the market. And these have only
increased between 2013 and 2016.
So the only explanation for why the latest BIS report put the FX
market at $200 billion less than three years ago is that the
market structure has changed to artificially constrain volumes.
And one of the factors that has artificially constrained volumes
was emphasised by the SNB Day, and that is risk management.
The SNB was a market event, and while it was not an operational
risk event it had a knock-on effect for pre-trade risk, which is why
we’ve seen credit conditions become more challenging.
GS: You’re referring to the fact that a number of FXPBs have
either reduced their client headcount or stopped offering this
service completely. Do you see this trend continuing or will
they return as new technologies like compression or
blockchain re-balance the economics of these businesses?
DJ: I think that they can definitely come back to the market
because the reasons why credit became artificially constrained
are solvable. PB is going through a transition. It’s traditionally
been heavily concentrated on looking at the credit name, then
doing due diligence, looking at the margining of the client and
assessing the market risk. But the biggest piece of this is now
operational risk, it’s calculating in real-time at any given
moment where your clients’ risk is and how much of the bank’s
balance sheet they’re using.
The operational risk is the only piece that you can actually
Yes, you can reduce market risk by having data analysis
regarding outlier events, and yes, you can always have a slightly
more rigorous due diligence process. But the thing that you can
really control is operational risk by adding technology.
Right now FXPBs are transitioning from being a back-office post-trade
function to a real-time pre-trade function, and technology
will enable them to do this. This is actually an opportunity to
unlock a tremendous amount of value for PBs and bring back a
good chunk of that lost $200 billion per day in FX ADV.
GS: If the FX market does start growing again and attracts
new participants from other markets, will the level of
fragmentation that we’ve seen prove unsustainable?
DJ: We’ve seen the first sign of consolidation in probably eight to
10 years with the closure of Bloomberg Tradebook [for FX], but
during this time how many hedge funds have started hanging
out a shingle and claiming to be market makers? This is the
excitement of an OTC market: every trader becomes a venue and
yet connecting all of them isn’t hard because of technology.
I think that a good, high performance, aggregator is the cost of
entry, but it needs to have a historical data set.
done some studies looking at fragmentation in the FX market,
and we found that while there used to be five venues required
to trade the euro, now people are trading on 16 venues to find
liquidity and to minimise market impact. For the peso, firms
used to have to trade on five venues, now it’s nine. So in
general, fragmentation is constant and will remain so,
especially with volatility. The only exception where liquidity
becomes more concentrated is during known and planned for
risk events like Brexit or the US elections, otherwise liquidity is
actually spreading out and becoming more fragmented during
GS: In general, would you say that trading in the FX market
has become harder as a result of this fragmentation?
DJ: It’s more complicated but firms have more tools at their
disposal. In the end, coming up with a great trading strategy is
still as easy or difficult as it’s ever been. Put simply, everything
has just gotten faster.
GS: But does this idea of everything getting faster just lead us back
to a latency arms race? It seemed like we were past that in FX…
DJ: The next stage of the latency arms race centres around
manual traders. It’s about providing manual traders with tools
of convenience so that they can press a button and execute at
speed, and then also constantly re-submit orders at speed.
The truth of the matter is that the real war is being fought on
the server side, not the GUI. By the time that a price gets from
LD4 or NY4 to a user interface, it’s already historical market
The maximum number of times that a human eye can
register a price change and understand what it means is three
times per second, but meanwhile the quotes in the market are
changing hundreds of times per second.
So you’re only getting a snapshot of those changes on the GUI,
so everything that you’re seeing is indicative and meanwhile
there’s still a good chunk of the market that is manual and
voice driven. Of the $1.7 trillion per day spot market, $500 billion is voice, which requires looking at a screen and then
trading. We have to empower these traders with tools that
enable them to trade algorithmically and simplify their lives so
that they’re not constantly re-submitting prices.
GS: Can you be more specific about the types of tools that
these traders need?
DJ: They need a quantitative, almost data science driven,
analysis and understanding of their counterparties. They also
need tools to ensure better neutral execution quality, so they
don’t have to do the work that computers do of breaking up
orders, and they can send large sizes into the market without
moving it and with minimal market impact.
GS: Is the increasing automation in FX and the increasing use
of technology an unambiguously good thing for the industry?
DJ: Well the technology and the code itself is always written by
humans, it always comes back to humans.
GS: Right, but what about things like flash crashes? In a world
of news reading algos there is more possibility of sudden
market movements triggered unexpectedly.
DJ: If you look at the sterling flash crash, it’s a perfect example
of us as an industry having to make assumptions because no
one has a complete snapshot of what happened in the
microstructure of the market.
Right now there’s unconfirmed rumours about whether it was
caused by a barrier options trader or by a fat finger and the
Bank of England and BIS haven’t been able to attribute it to
one factor. The fact that in the largest market in the world no
one out there has the data set to really analyse this is
something of an irony.
We’re actually using data to conduct our own analysis. Our goal is
to eventually be able to do a post mortem for any flash crash within
24 hours and then at some point in the future send early warning
signals about characteristics seen in other flash crashes.
An interesting point to note about the pound flash crash is that it
occurred at a time when the currency is not normally that liquid,
and yet in the two hours preceding the crash it was actually
So it was trading at an above average volume,
and then dropped 200 pips in 10 seconds, that’s massive.
But if we’re going to make the hypothesis that algo trading is
driving a lot of these sudden moves then we should also make
the equal assumption that those 10 seconds represent a lifetime
for algos to get out. So we can’t be certain that it was algos that
caused it, we don’t know for certain what happened yet.
GS: Would circuit breakers help alleviate flash events?
DJ: Actually, if you look at the pound flash crash, there were
circuit breakers in place at the CME. But you have to remember
that there are all these prop funds that are sitting between the
CME and NY4 and LD4 and TY3 and they’re constantly market
making between the spot and the futures markets.
So when the drop in the pound triggered a circuit breaker and
stopped trading on the CME, if anything this would exacerbate
the liquidity problems because one of these firms’ main venues
for price discovery has stopped, but meanwhile we don’t have
circuit breakers in FX. If the CME didn’t have circuit breakers
then perhaps more of these market makers would have been
there at every level and perhaps the price wouldn’t have
gapped so dramatically.
GS: You mentioned transparency earlier and you’ve referenced
the need for the greater availability of data in FX a couple of
times. I don’t want to get into too much detail, but I feel like I
ought to ask: how is the consolidated tape developing?
DJ: It’s going well. Consolidating and multicasting the data in
microseconds represents a huge software performance
challenge, but fortunately this is the exact challenge that we’ve
spent the past five years developing technology to overcome.
The early adopters of the tape have mainly been on the buy
side, but I think that’s okay because we see the FX market
moving towards a disclosed liquidity model and a lot of these
buy side firms that are market making directly to firms in the
market. Then behind them there are some ECNs that are
interested and potentially a few banks.