When things go wrong in automated high frequency trading, they can go disastrously wrong. This is perhaps unsurprising when you consider that about 165,000 separate trades can be completed by a high frequency trading platform, which buys and sells large volumes of stock in short periods of time, in the time it takes for Usain Bolt to react to a starting pistol1.
One such example was the so called "flash crash" on 6th May 2010, where the Dow Jones Industrial Average plummeted 600 points in 5 minutes losing 7% of its value. Approximately $1 trillion in market value disappeared temporarily, only to recover those losses in the following 20 minutes, which serves to highlight the magnitude of market failures exacerbated by high frequency traders and trading algorithms.
The recent IT failure at Knight Capital reminds us that automated high frequency trading can cause significant market risk and no one knows who will be next to suffer such a catastrophic event.
What went wrong at Knight Capital?
The headlines were that there was a defect in a high-frequency trading algorithm that lasted for 45 minutes which generated thousands of mistaken orders for approximately 150 stocks (buying stocks at a high price and selling stock at a low price). When Knight Capital sold out of their positions they were left with a $440 million loss. The net effect resulted in Knight Capital's share price to plummet 75% of its value in two days and pushed Knight Capital into seeking emergency funding by a group of independent investors in exchange for a 73% stake in the company.
It is thought that the trading program error was linked to the launch of the New York Stock Exchange's "Retail Liquidity Program" (RLP) on 1st August 2012. The RLP was designed to offer individual investors the best possible bid-offer spread. This meant market participants had a relatively short period of time (under two months) to write computer code to make use of RLP. The trading program error at Knight Capital happened in the first 45 minutes of trading on 1st August 2012. Although Knight Capital has not been forthcoming with detailed explanations as to what precisely went wrong, the prevailing view of independent commentators is that the computer code that went into the live environment to introduce the RLP was:
- actually a test program designed to simulate trade requests and evaluate if they went through properly2; or
- not exhaustively tested and failed to pick up the interaction with a high-frequency trading algorithm3.
What are the effects?
First and foremost there are effects on the company affected by the trading glitch on their reputation in the market place and often their financial viability which causes clients to drain away. It has been a very embarrassing episode in Knight Capital's history as it had previously been a well-respected market maker until the incident. In the immediate aftermath Vanguard Group and Fidelity Investments did not route trades to Knight Capital due to a lack of confidence in Knight Capital's systems.
There is also a wider macro effect as a result of lower confidence in the strength of the market infrastructure to deal with such disruptions which lead to large losses. A recent study showed that only 2% of 260 respondents rated their confidence in the US equity market structure as "very high" compared with 12% after the "flash crash" in 20104. The likelihood is that pending improved regulation in the area of automated high frequency trading in the UK and US is likely to bolster the low confidence in the market.
It is understood that the Financial Services Authority (FSA) in the UK is considering introducing stricter rules in relation to high frequency trading following a review of such practices earlier this year5. So it is highly likely that further regulation in this area will also be introduced in the UK as well as the US following the Knight Capital crash.
Prevention is far better than cure
As lawyers in the financial services field we scratch our heads and wonder what part we can play in protecting our client or our company from such a risk. There will certainly be a role in interpreting the forthcoming regulations regarding high frequency trading to be introduced by the FSA. In addition, to the extent that there are any IT suppliers or external consultants involved in providing program code for trading purposes lawyers acting for the recipient will need to ensure that as much risk as is reasonably possible is placed on the IT supplier to deliver suitable and tested software code and implement such code in a way that minimises the risk of a trading collapse.
It is customary in important services agreements to have a disaster recovery plan if, for example, unusual incidents beyond the reasonable control of the supplier (e.g. fire or flood) occur without warning. The impact on the provision of the services is reduced by the implementation of a well thought out plan as to how to deal with such circumstances to ensure the continuation of services. If trading companies have not done so already they should consider having a financial disaster recovery plan, such that certain steps are identified in advance which would be taken in the event of a catastrophic financial incident (such as a defect in a high frequency trading platform) occurring.
There are also practical and operational issues which the legal team can raise in an effort to de-risk the business of a trading company, such as:
- Testing code. Clearly had the defective software code been tested prior to introduction onto Knight Capital's live IT environment then it is likely the disastrous ramifications would have been picked up nullifying the problem. In addition, any such implementation of new software code could be subject to a soft launch in the live environment prior to market opening (e.g. at a weekend or overnight) and monitored so that any unusual activity can be identified at the earliest opportunity.
- Human oversight. A criticism of Knight Capital is that
it took so long for them to notice the problem when other market
participants were aware of a potential issue minutes after the
market opened6. Stating the obvious, trading companies
should ensure that there is a human being which is responsible for
monitoring all trading systems of a company at times when the
markets are active to identify irregular trading emanating from its
automated trading systems. Although such algorithms have their own
in built safety systems human oversight can double-check when those
are not performing or there are unexpected situations. Clearly the
nominated person should have written operating parameters for each
high-frequency trading system and where a system performs outside
such parameters have the authority and instant ability to flip the
"off" switch to give the company time to react and
rectify the issue.
- Circuit breakers. Following the "flash crash" the Securities and Exchange Commission (SEC) introduced circuit breakers based on the increase or decrease in the price of individual stocks. Perhaps there also needs to be a wide spread introduction of circuit breakers in relation to large trading volumes on particular stocks as the price based circuit breakers didn't avert the large scale losses of Knight Capital. The SEC is looking at this issue and will "accelerate ongoing efforts to propose a rule to require exchanges and other market centers to have specific programs in place to ensure the capacity and integrity of their systems"7.
What can be done if the worst happens?
It is clear that regulators (particularly in the US) will be reluctant to help. The SEC refused to have the majority of the $7 billion worth of trades made by Knight Capital cancelled as those trades didn't increase the share price of the majority of the purchased stocks by more than 30%, the cancellation threshold, which followed the rules established after the flash crash in May 2010. This reiterated that regulators are not there to keep financial institutions afloat but rather guard against systemic risk in financial markets.
Even though legal terms may state that a trading company is not responsible for trading losses of its clients, it is unlikely that a company would choose to place the losses on their clients as it would destroy its client base and it would be unlikely to receive any support in the market to recapitalise itself.
Without a shadow of a doubt, prevention is better than cure.
Closing thoughts
Knight Capital had become an important broker in US stock and accounted for approximately 10% of daily US trading volume, which resulted in it receiving 90 or so offers to recapitalise following the IT failure8. However, other less prominent market participants are not so likely to be as lucky and should consider both ways to prevent such issues arising in the first place and a financial disaster recovery plan if they do.
In the same way companies spend vast amounts of money for ever greater transaction speed and lower latency, perhaps they should also spend just as much money (if not more if they are interested in preserving their business for shareholders) on reliable order stopping/capping software to prevent a repeat of the Knight Capital crash. Coupled with taking the prudent measures set out in this article in advance of an incident, this will ensure that trading companies are doing their utmost to de-risk their business.
Footnotes
1 http://www.bbc.co.uk/news/magazine-19214294
2 http://www.cio.com/article/713628/Software_Testing_Lessons_Learned_From_Knight_Capital_Fiasco
3 http://www.businessweek.com/news/2012-08-14/knight-software
4 http://www.ft.com/cms/s/0/20c8c082-eba5-11e1-985a-00144feab49a.html#axzz24Ib5cPka
5 http://www.marketwatch.com/story/uk-considers-more-high-frequency-trade-rules-2012-08-13
6 http://dealbook.nytimes.com/2012/08/03/trading-program-ran-amok-with-no-off-switch/
7 http://www.sec.gov/news/press/2012/2012-151.htm
Thomas Joyce CEO of Knight Capital on the IT glitch: http://video.cnbc.com/gallery/?video=3000107449
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.