Handing greater oversight of algorithms to regulators may help identify the causes of market dislocations, according to responses to May's poll on theTRADEnews.com, but many market participants are worried about the impact on execution performance.
The ability to identify more effectively causes of future market disruptions was voted by 41% of readers as the most significant impact of stress testing of algorithms, with 38% voicing fears that any such measures would damage execution performance. The remaining 21% asserted that vetting algorithms would reduce the chances of a repeat of the flash crash of 6 May 2010, a brief but severe downturn across US markets that was instigated by a large trade entered by a institutional investor via an algorithm.
But while the majority of poll respondents thought that algo testing would have some positive impact on market stability, some buy-side traders are convinced that it is not that algos need to be monitored.
“The algo is seen as a panacea, but it is not,” asserts Tony Whalley, head of dealing and derivatives at Scottish Widows Investment Partnership. “Algos are merely one tool in the execution kit available to the buy-side trader, so it is the trader that needs to have the ability to pick the right tool for each order.”
Moreover, the vetting of algos by regulators may prove impractical for a multitude of reasons.
“If an algo is put through a certification process, it may slow down the time to market and could therefore raise some competitive issues in terms of time to market,” says Brian Schwieger, managing director, head of EMEA algorithmic trading at Bank of America Merrill Lynch.
Schwieger also notes that an algo's performance in one market won't necessarily be replicated in another, thus adding another layer of complexity to the vetting process if checks need to be applied across asset classes or trading venues.
Market participants suggest that focusing efforts on circuit breakers would provide a much more consistent way of preventing erroneous and potentially damaging trades.
“Circuit breakers will prevent algos going wild and a market-wide solution would cover everyone,” adds Whalley. “Algos are electronic tools and cannot think for themselves, which means they may encounter market conditions that haven't been programmed in. Is that the algo's fault or is the trader that hasn't placed any limits on his order to blame?”
Schwieger believes that systems already used internally by brokers could be extended. At Bank of America Merrill Lynch, each algorithmic order sent by a buy-side client is assessed against a profile to determine whether the instrument being traded is appropriate for the chosen algorithm. If the order is deemed unsuitable, a BAML sales trader will call the buy-side trader to explain.
“You could apply this market-wide and check orders for factors like volatility spread, trade frequency and average trade size,” said Schwieger. “Arguably this would be more efficient. In other words, instead of certifying algos, markets could be certified as an alternative in terms of their suitability for algo use.”
Policy makers from all corners of the globe are currently discussing ways to better manage algorithms as their use grows and strategies become more sophisticated. A handful of Asian markets, namely Thailand, Indonesia and India – where electronic trading is still a growing trend – have already implemented prior testing of algorithms before they can be deployed. US regulators have focused mainly on pre-trade risk controls at the broker level and harmonisation of circuit breakers for trading venues, while Europe is weighing up its options for approving algos via the MiFID review.