Fail-safe software could stop flash crashes

HIGH-FREQUENCY trading algorithms are seriously profitable. But they are also a serious problem, leading to mysterious "flash crashes" on the world's financial markets. So would emergency fail-safes of the kind used to prevent medical robots and nuclear reactors going haywire be any help?
Philip Bond, a computer scientist at the University of Bristol, UK, thinks so. At present, suspicious trading at an exchange can be stopped by a catch-all "market halt", but that's often too late: what is needed is a reliable way of sensing when wild stock-price variations suggest a build-up to a crash. At that point a smart circuit breaker could step in, Bond told a London meeting of the Centre for the Study of Financial Innovation on 11 December.
The idea is to avoid another debacle like the flash crash of 6 May 2010, when $10 trillion was briefly knocked off the Dow Jones Industrial Average after a trading firm's high-frequency algorithm went awry. Prices mostly recovered in a matter of minutes, but the turbulence caused slumps in stock prices around the globe.
High-speed traders have a distinct advantage over traditional investors, according to research by Andrei Kirilenko, chief economist at the US Commodity Trading Futures Commission, and his colleagues. Kirilenko presented the findings on 30 November at the National Bureau of Economic Research's Market Microstructure Meeting. This skewed playing field might increase market instability even more.
Because high-frequency trading algorithms (HFTs) can work at blistering speeds - a trade every 60 microseconds is common - a lot can happen before humans have a chance to intervene. This has left governments scrambling to come up with ways to regulate the practice. The European Parliament is considering legislation that could force traders to increase trading intervals to a "safer" half a second.
But Bond and his colleagues, who examined the risks of high-speed trading for the UK government, say that enforcing a delay is wrong-headed. Making algorithms wait half a second would stop them from reacting to breaking financial news and render them useless.
Instead, they propose using circuit-breaker algorithms that will trip when trading becomes erratic. Such software monitors systems like medical robots and nuclear reactors for potentially dangerous behaviour. "HFT circuit breakers need to be considered as very high reliability software engineered to work under stressed conditions and with multiple backups," he says.
A decision is needed fast. The proportion of high frequency trades has declined in the US in recent years as they have lost some of their initial edge, but they still account for a majority of trades. And the hardware needed to get in the game is becoming more accessible - a server running a high-speed system costs only $270,000, and prices will keep falling.
Any technical backstops will have to distinguish between algorithms gone haywire and simple bad decision-making, warns Fod Barnes, an economist with UK consultants Oxera. "The HFT system is a bit more complex than just engineering," he says. "Why should those who manage to write an algorithm that makes a series of very bad trades be protected from their own folly?"
 
Author: Paul Marks
Source:

0 yorum: