A-Team Group’s London TradingTech Summit on February 27, 2019 hosted a lively discussion on Post-MiFID II use of algorithms and smart order routers (SORs). Gavin Williamson, former managing director, global head of product development and regulatory change at HSBC, moderated the debate that ranged from total cost analysis (TCA) and pre- and post-trade analytics to machine learning, algo wheels and data management challenges.Speakers included Joe Everitt, managing director of electronic trading, SELECT at Stifel Electronic Trading (Europe); Will Winzor Saile, execution analytics and architecture at Redburn; Peter Simpson, vice president of Panopticon Streaming Analytics at Altair Engineering; and Jonathan Finney, director of European business development at Citadel Securities.
Post-MiFID II, the buy-side has been under intense pressure to evidence best execution in its reporting. With the Financial Conduct Authority (FCA) strengthening its approach to risk management and expanding into areas such as governance and market conduct, the industry approach to risk controls has evolved, with an inevitable impact on the design, deployment and development of algos.
One panel member explained: “We have responsibility for managing relationships with vendors and outsourced algo providers, and under MiFID II there is no way you can offset that. You can’t outsource the responsibility – you need to know the systems you are using, how you are using them, and how your clients are interacting with them.” However, not everyone agreed. “Nothing has really changed,” noted another panel member. “We are doing exactly what we did before. But having said that, anyone pre-MiFID II who was running an algo model without risk management, without kill switches etc, was doing it wrong. All we are doing now is running a record – it is about having the documentation to evidence that we have these capabilities, rather than implementing them, because to be honest we had them all before.”
There has certainly been an increase in monitoring, not just around how well a trade has performed, but also around how orders are placed in the market. There has also been a move towards more detailed live analysis of each trade’s journey. As the onus transfers back to the buy-side in terms of monitoring this, the algos themselves have become steadily less complicated. Conversely, smart order routers (SORs) are becoming more complex.
“We are finding in the post-MiFID II landscape that liquidity has fragmented. SORs are under increasing pressure to adapt to this new scenario,” said one panel member. “There are so many new venues out there, and anyone trying to operate a machine learning SOR based on a couple of weeks-worth of data is not going to do very well,” agreed another. “If you are trying to execute a trade, based on limited data, it is difficult for a machine to know what to do. But how much data is enough? The buy-side will always worry about how much data is needed to be significant.”
As well as the algo and SOR issues resulting from MiFID II, there are concerns that a lot of the data introduced by the regulation is not as effective as it could be. “Published information about trading performance is generally out of data, inaccurate, or obscured so that competitors can’t use it,” said panellist. “When key outputs from MiFID II don’t provide any useful insights, it is hard to justify the additional effort they require. This is simply not information that people want to share.”
Overall, MiFID II has had an enormous impact, both on the implementation, usage, application and selection of algos and on SORs and their suppliers. Going forward, one panel member suggested: “Algos will be a lot more targeted, In the past few years, there has been a trend for the sell-side to innovate in terms of coming up with a new algo, giving it a funny name, then finding a buy-side trader who likes the sound of it and will give it a go. That kind of thing can’t happen anymore. The process is much more pared down – you need to hit this target, under these circumstances. Algos will be much more about how we can deliver specific goals in the best way, rather than coming up with new ways of selling and marketing them.” “The onus is on the buy-side, and it will become more and more responsible for the actual instruction of the algo,” concluded a panel speaker. “By the end of the year, I think we will have more uniformity around that, and we will start to see some real differences between brokers on the back of that development.”