Two of the most expensive losses in the history of the
banking industry were definitely caused by humans:
I. Nick Leeson, a British, Singapore-based trader,
personally defaulted Barings bank with a combination of:
·
speculative trading
with large sums of money on the presumed direction of the Nikkei index;
·
hiding the mounting
losses on a secret account out of sight from his peers and superiors;
·
trying to make up
for earlier losses, by executing the one gamble that could wipe out all
previous losses, which of course… it didn’t: the Las Vegas-syndrome;
Starting in 1992 with his speculative actions, Nick Leeson caused Barings
a mindboggling damage of $1.4 bln, when he eventually fled the bank at broad
daylight in 1995. At the time, Barings was finished as a private bank and was
taken over by the Dutch banking conglomerate ING Groep NV
II. Jérôme Kerviel, an unsuccessful French trader that
was dissatisfied with the size of his bonuses, tried to pull the same stunt at the
French bank Société Générale ($SCGLY) as
Leeson did at Barings:
·
Kerviel speculated
with large sums of bank money (north of €50 bln) on the presumed direction of
the DAXX, Eurostoxx and FTSE indexes;
·
He created a fake
company that would operate as a ‘straw man’ for his orders;
·
His main purpose
was not to peculate his profits from the bank (according to nl.Wikipedia.org). Rather to the contrary, he wanted to use those profits for improving his trading results, in order to become a
more successful trader and receive a higher bonus in the process;
Although Kerviel’s fraud only lasted for a few months (from the fall of
2007 until the spring of 2008), the results of it were beyond imagination: a
direct loss of €1.5 bln and subsequent losses of €3.4 in the process of unwinding
the open contracts that Kerviel’s actions left behind. This was officially the
biggest fraud case in the history of banking. Only direct intervention of the French
government could supposedly prevent SocGen from defaulting in 2008.
These were kinds of fraud previously unheard of in
their sheer consequences and the financial havoc they caused in the banking
industry. As stated before, Barings ceased to exist as an independent business
bank and SocGen suffered also gravely from the consequences.
Still, these cases of fraud paled in comparison to
what happened last week: an error in a computerized High Frequency Trading
system of Knight Capital Group (KCG
) caused this bank to lose $440 mln in … just 45 minutes!
The absolute amount of the damage caused by the events
at Knight Capital is not extremely large (“only” $440 mln), when compared to
the previously discussed cases of fraud. What is flabbergasting, however, is
the speed in which the events took place and the displayed impotence of the human
controllers against errors in the HFT computer systems: they stood by and
watched the havoc unfold!
Here are the pertinent snips of an FT article describing
the events that took place at Knight:
It took just 45 minutes for a trading glitch to create a $440m loss for
Knight Capital this week.
That loss is almost double the revenue generated by the company in its
entire second quarter and was enough to wipe three-quarters off the value of
the market-maker within 24 hours. For Wall Street, it’s the equivalent of
watching a state-of-the-art aeroplane come crashing to earth after a momentary
software problem.
Last year, Knight dominated trading in NYSE and Nasdaq-listed stocks and
was routinely showered with awards for its advanced information technology and
high-speed trading systems.
Knight is now said to be looking for a partner to help shore up its
capital and possibly even rescue the company. It’s a stunning defeat for a
market-maker that has come to embody the silicone-coated transformation of
modern trading.
The question now is whether this series of
technical troubles will force financial companies to rethink their use of
technology, even after trade “automation” has become the buzzword of Wall
Street in the face of new profit-crimping regulation.
Credit Suisse analysts recently estimated that pure
“high-frequency trading”, which includes algorithmic strategies, now accounts
for almost half of US stock trading. “Real money” stock trading is now at
decade lows, they say.
The kind of algorithmic trading employed by Knight Capital thrives on
automation; hire a skilled maths expert to create a complex formula that
arbitrages minute discrepancies in stock prices, ensure that it’s executed at
speeds measured in milliseconds, set it loose in markets and sit back and reap
the profits.
For Knight, that automation went spectacularly wrong on Wednesday. The
company said it installed new market-making software that ended up “sending
numerous erroneous orders in NYSE-listed securities into the market”.
To market-watchers, it looks as though the software was spitting out
orders for stocks and then buying them up almost instantaneously, a pattern
which, as market data firm Nanex points out, would mean Knight instantly losing
the difference in price.
Years into what’s been described as a “high-frequency arms race” among financials, these companies are reaching the limits of available
technology and pushing what they have to greater extremes.
This article in the Financial Times brings me to the explanation
of the title of my article. The combined losses of the damage caused by Leeson
and Kerviel are more than ten times as high as the damage of the events at
Knight Capital.
The difference is, however, that normal banking controls
and regulations, a little bit less trust and more ‘spider sense’ among the
managers and peers of these two gentlemen would almost certainly have prevented
these events from happening. These guys were humans that acted very human and almost
got away with it, because their direct colleagues and management acted very human too. In
optimistic times nobody sees a raincloud until it starts to rain.
The scary part at Knight Capital was that no human
could have prevented these events from happening, without abolishing the
HFT-systems as a whole. Understanding what happens in the HFT code during critical
situations is much to difficult for normal humans, especially in the very
detailed way that is required to thoroughly test this software. People trust
the mathematical magicians, simply because they don’t have the power to prove
them wrong.
The Financial Times does a nice job with this article
and the question it asks (first red paragraph) is a good one. However, there are
two much more fundamental questions that ought to be answered first.
·
Should we implement technology that we fundamentally don’t
understand and that we can’t possibly check on being flawed anymore ?
·
Should we implement technology that we can’t control
anymore with our human senses and that can create havoc within milliseconds?
Everybody that saw the true purpose
of ‘Skynet’ unfold in the surprisingly good part III of The Terminator
franchise, knows that these are the most fundamental questions concerning
computer technology. Especially in a time where our lives are increasingly
interconnected with computer systems and we increasingly lose influence on the
quality and maintainability of the computer code in these systems.
Hence, when only 500 of the ca. 6,000,000 lines of
computer code in a Boeing 777 contain (fundamental) flaws, then Boeing’s software
is 99,99167% error free. On the other
hand, would you be the one to find out what the errors in these 500 lines are all
about while being in mid-air at 30,000 feet of altitude? The hell you wouldn’t!
And would you like to see your
life-savings vaporize at a bank or an insurance company, due to an error that
some nerd programmed in a mathematical formula that nobody else in the company
fully understood?!
I am a vastly experienced software-tester
with more than 14 years of hand-on experience and – I may say – a darn good one. Today, I had
another successful implementation of new software at the global system bank
where I work.
However, if someone asked me to test
and take the responsibility for the HFT-software that Knight Capital
implemented, I would reply that I’m not smart and intelligent enough to do this
job and take this responsibility. Seemingly, I was not the only one that should
have said this.
The failure at Knight Capital cannot
be blamed on the nerd that programmed the mathematical algorithm, although this
will definitely happen in the near future. The true failure lies by the people
that don’t understand the ultimate consequences of the two very fundamentals
questions that I asked. Both questions must be answered with a clear “no”!
