Christopher Clack, Department of Computer Science, University College London
"HFT algorithms are more sensitive to small delays that happen more often."
London - Is the future ready for new technology? That was the question posed to panellists considering developments in automated markets at FIA's IDX conference yesterday. The question is being asked against a backdrop of major regulatory overhaul by European markets regulator ESMA, currently in consultation and discussion with the industry.
The debate in Europe, said Sonia Cattarinussi, director of the market regulation division at France's markets regulator AMF, is getting more mature as rules are written that meet the needs of many market players.
"What we have been trying to do at ESMA level is to set up minimum sets of rules where the objective is to secure the environment," she said. "The mandate given to ESMA to tackle these issues, its focus is on market quality improvement."
From a market actor perspective, she said, that is about having internal controls for algos as well as clear chains of responsibility. On the exchange side, rules are intended to ensure there's the capacity to handle, for example, high volume messaging.
"What we focus on is trying to implement the appropriate tools which will improve orderly trading," she said. "For us, the debate is - can all market actors given the evolution of technology on markets still interact (with) each other and contribute to the price formation process?"
The 2010 "flash crash" in the US still looms large in the minds of regulators. Though there has not been a pan-European event of similar scale, the industry is not willing to take it for granted that there won't be.
Christopher Clack, founder of Financial Computing MSc programme at University College London presented some preliminary research findings into how algos trade and interfere with each other. Algorithms can become "locked in a deadly embrace" where they become unified. These coupling chains can loop back on themselves to form feedback loops.
His team found a number of triggers for feedback loops and one of them - unexpected delays in information - affects HFTs more than low latency or slower traders. He explained further that "HFT algorithms are more sensitive to small delays that happen more often".
Those kinds of findings further bolster the idea that market data distribution is crucial to orderly markets.
"(It) is very important from a risk management perspective that everyone knows, (for example) this price maybe should be on a millisecond order match. I got it at 100 milliseconds, or 2 seconds late, so something must be wrong," said Wolfgang Eholzer, director head of section Trading System Design at Eurex. "Then I can take corrective actions: I can stop my algorithm, pull my orders, send an alert to my trader depending on the trading strategy."
He added: "We are very committed about sharing that information in the most efficient way with the community."
Christian Hauff, CEO and co-founder of Quantitative Brokers, said he is comforted by circuit breakers and increased regulatory auditing to identify excessive actions. At the same time, on a practical level some considerations need to get made for the wide variety of trading strategies and how they operate.
"From a buy side agency algorithmic perspective, we have other overriding constraints and obligations on that order which may make our interactions more irregular and maybe less predictive," he said.
Some of the issues rest with the perception that markets are widely rigged against non-speed traders. Combatting such perceptions means getting to grips with how transparency should function, said Stephane Tyc, co-manager of Quincy Data and McKay Brothers.
"How do we publish the right amount of data so that everyone can analyse what is going on?" he said. "(Real-time data) doesn't help the end user really understand what went on…(it) helps executing brokers, high frequency traders but (not) the end user understand that this order was well managed."
He applauds ESMA's move to synchronise data as part of its MiFID II reforms, though he thinks it should include precision. "Under 10 microseconds would be good," he said.
After MiFID II is implemented, panellists believed hot topics will be transparency and data analyses, liquidity provision and management, and latency determinism.