On 17 August, a day of high trading volume and volatility, a number of smart order routing providers suffered outages and other difficulties, disrupting trading for some of their clients. “During that week, we and a number of other providers did not have the level of capacity that we needed,” says Richard Evans, global head, Citi Electronic Execution, which has responsibility for Citi’s smart order router Lava. However, it is exactly at these critical times – high-volume, high-traffic days – when traders need this technology the most and cannot afford these kinds of outages, according to one well-placed industry source. “These disruptions defeat the purpose of having such technology in the first place,” remarks the source. “Over the last several weeks, when we have had days of huge trading volume and volatility, the smart order routers have not been able to keep up,” he adds.
“Due to expected high volume today we may see delayed market data feeds from our vendors,” read a Bear Stearns algorithmic trading alert sent to the firm’s smart order router (SOR) clients on 17 August. “This may affect our SOR Stealth orders. As a pre-emptive measure please do not send orders to SOR Stealth. Use SOR Post, SOR Reserve or SOR Cloak,” it continued. “We have had another front door issue causing most of our clients to disconnect and not be able to reconnect,” read an email to Lava’s clients from Citi Global Market’s Electronic Execution department. There was a downtime of about 45 minutes for a limited number of clients who encountered difficulties accessing the Lava front-end. Goldman Sachs Algorithmic Trading (GSAT) is also rumoured to have encountered difficulties, although the firm declined to comment.
What was the cause of these problems? “The problem lies not with Lava’s scalability, but with dramatic surges in market data, which leaves us slightly ‘caught on the hop’,” comments Evans. These issues, when they occur, are short-term, never lasting more than 24 hours, he says.
According to Matthew Celebuski, senior managing director, Bear Stearns, much of the problem lies with exchanges. “After the first phase of Reg NMS was implemented in March (when all market centres under the jurisdiction of Reg NMS instituted trade-through protection rules under Rule 611), some exchanges had problems in getting us timely data,” comments Celebuski. “You would see one exchange slow down and then another.” Very high volumes of market data, some of which is delayed, affect the underlying algorithm’s ability to function.
Under Reg NMS, these issues become a major concern. Under rule 611(b), the so-called ‘self-help’ exception, trading venues are permitted to execute trades on one market centre at prices worse than those quoted on another, if the latter is taking too long to fill incoming orders. A delay of more than one second in responding to an incoming order is deemed too long under Reg NMS and gives brokers the right to declare ‘self-help’ against the exchange in question, excluding it from its venue selection process.
Technology failure has important implications for brokers’ clients under Reg NMS. “How do you know that you are indeed executing where you should be and you are not getting ‘traded through’ on another exchange?” asks theTRADEnews.com’s industry source.