Algo Testing Under the Regulatory Microscope

Algorithmic trading is not as sexy as author Michael Lewis would have you believe. In fact, algorithms are really quite a familiar and accepted part of modern life. Just like in ordinary retail experiences, algorithms can be used for mundane associations based on other customers and previous purchases, or even instantly producing “Keep Calm & Read Waters” t-shirts advertised in the window of my search engine, based on my Google history. 

In the finance world, the role of algorithms has expanded exponentially. At their simplest, algos speed things up, increase liquidity, and can employ hugely complex artificial intelligence (AI) reasoning to process staggering amounts of incoming data. Algorithms also don’t have feelings, they don’t need coffee to operate, and they don’t take holidays.

However, that’s not to say that algos are faultless—far from it. Over the last few years, algorithms have become the focus of a significant amount of negative press. It’s mostly misunderstanding, but in an era when the finance industry—especially after the global financial crisis of 2008—does not endear itself to the public imagination, the idea that there could be a shadowy computer manipulating retirement portfolios presents both politicians and the press with an opportunity to raise public awareness and play the game to their advantage.

As such, and for the first time in Europe, Mifid II will present a regulatory framework for algorithmic trading with new timestamping, testing and best-execution requirements. But is this like taking a sledgehammer to swat a fly? What are the changes going to mean for capital markets participants, and are these actually much-needed regulations that might have a positive effect on transparency, trust and the public relations battle that algorithmic trading finds itself fighting in the mainstream media?

Demystifying the Algo Trader

“Mifid II is throwing ropes around the horse and tying it to a fence post. With Mifid II, they’re aiming to increase transparency and to maintain an orderly market. And to do that, they’ve thrown ropes around high-frequency traders to rein them in.” Terry Keene, Integration Systems

There are three basic types of algorithmic trading. The terms are used synonymously and they have much in common, but they’re also subtlety different, and distinguishing them helps us understand where transparency is needed and why.

Terry Keene, CEO of Integration Systems explains: “The first type is electronic trading, sometimes called algorithmic trading. That’s a pretty broad brush, but it refers to a firm that trades using electronic algorithms. For example, if I was a day trader with a stop loss on a stock, one might really stretch that and say that it’s algorithmic because the computer is putting the sell order in when it hits my stop.”

The second category, according to Keene, refers to market-makers. “These are algorithmic traders but they don’t really depend on speed for their algos to be successful,” he says. “They are paid primarily by venues to make markets and create liquidity. They hope to make their money on transaction fees, and maybe a little bit on the spread.” 

For market-makers, the job is to get in and offer buy and sell opportunities continuously so that when traders enter an execution venue to buy or sell their own stock there is someone on the other side of the trade to ensure execution. It’s not terribly sophisticated—it’s a bit like a game of algorithmic musical chairs and it keeps the markets liquid and functioning as they should. 

The final algorithmic trading category, according to Keene, is where things start to get interesting: high-frequency traders (HFTs), the subject of Flash Boys, the supposed culprits of the May 2010 Flash Crash. When asked if Mifid II might kill or dampen HFT and algorithmic trading, Keene says, “These are the smartest guys on the planet. A typical trading team consists of an old-time trader, a mathematics PhD quantitative analyst, and a couple of PhD computer scientists, who write the code. It won’t take them long at all to find out how to leverage the new regulations.” 

But will Mifid II curtail momentum in the HFT space? Not according to Keene. “I don’t think the regulation is going to stop or slow the algorithmic traders down. You might shift them a little bit, but no matter what you do they’ll figure out how to make a profit on it,” he says.

So how does one regulate the smartest guys on the planet? Is there an inclination to go in a tad strong? “Mifid was like letting the horse out of the barn. Mifid II is throwing ropes around the horse and tying it to a fence post,” Keene says. “With Mifid II, they’re aiming to increase transparency and to maintain an orderly market. And to do that, they’ve thrown ropes around high-frequency traders to rein them in.”

Mifid and Testing

So, for the first time, in the EU, rules on algorithmic trading will be imposed on investment firms and trading venues by way of Mifid II. The new regulation, as ever, is legislating for increased transparency—transparency, that is, for pricing, routing, execution strategies and other potential pitfalls in the journey between broker, exchange and investor.

In summary, what we’re talking about here, according to Giles Kenwright, head of regulatory Advisory at Delta Capita Consulting, is “pre- and post- trade transparency to a much more granular level of detail than is necessary today,” across a broader range of asset classes. Automated trading algorithms will also need to be tested to ensure sufficient controls are built in and clock synchronization will guarantee time stamping within 100-microsecond accuracy.

But the real shake-up will come with the new testing requirements, now inclusive of multi-venue testing for high-frequency trading. Old forms of back-testing algorithms will no longer suffice. Mifid II has very specific stress-testing requirements, but more importantly, if an HFT firm is trading arbitrage across multiple markets, multiple venues or even multiple asset classes, it is going to have to test those algos and demonstrate that they are able to maintain orderly market behavior across each and every one of those venues. Potentially, this is a great upheaval with intra-day tweaks becoming unfeasible due to the sheer scale of the regulation’s requirements. 

Speaking on a panel at the recent European Trading Architecture Summit 2015 in London, Tim Rowe, manager of trading venues policy at the UK’s primary regulator, the Financial Conduct Authority (FCA), spoke in great detail about the concerns, frustrations and hopes for the future of algo trading: “There is an assumption that the algo testing requirements should only apply to that complex, sexy stuff,” he explained. 

“In practice, it’s the simpler stuff that’s more problematic, when something is coded wrong or when someone gets a parameter wrong—and off it goes,” Rowe said. “Occasions when you see a whole stream of one-share trades going through the tape and you think, that’s the easiest algorithm in the world, right? But someone’s managed to mess it up. That’s why we say—test this stuff. And that’s why we have the systems and controls in place so that even when it goes live, there are controls so that if it goes wrong, it can’t go too far before something kills it.”

Misconceptions, Transparency and Trust

In an effort to further understand the positive implications the new regulations might have, Waters spoke to John Adam, global head of product management at London-based trading technology vendor Portware. According to Adam, the fear and mistrust that accompanies algorithmic trading exists between the public perception of a new technology that’s disrupting the existing way of doing things that can be used both well and poorly that is not well understood by the general public on one hand, and Portware’s clients, tier-one institutional asset managers, who see the true value in algorithms, on the other.

“If a head trader or CIO views algorithmic trading as a tool with which they can achieve efficiencies and economies of scale at their desk—and ultimately get a better trade for the end-user—then yes, they are going to do that and yes, they are going to advantage their desk by engaging with this kind of liquidity,” Adam explains.  

So are we in a better place to avoid the kind of disorderly market behavior that precipitated the Flash Crash and the fiasco that led to the decline of Knight Capital back in August 2012, when Knight deployed an untested algorithm in a live environment that ultimately led to a loss of some $460 million and the firm being acquired by Getco some five months later? 

“I think so, yes,” says David Masullo, head of EMEA sales at Bloomberg Tradebook. Masullo says there has been a general improvement in governance and technology. “Regulation is driving improved and enhanced governance, from a compliance standpoint. It’s not just about suspicious transaction reporting anymore—your overall controls and governance must include quantitative research that addresses the micro-structure of the markets in which you trade, as well as the algorithmic trading tools themselves,” he says. 

Best Execution

As important as wrapping up algorithmic trading in a regulatory framework is for providing greater transparency and restoring trust, it’s equally important to address the misconceptions that have led to this PR battle between mainstream news platforms and investment professionals.

Portware’s Adam says that it comes down to the quality of the algorithms themselves, the traders’ responsibility to achieve the best possible execution on behalf of the funds they represent, and the new regulatory direction that the industry is moving in. 

“It’s commendable that the Financial Conduct Authority has always taken a principles-based approach to this and I think that the Mifid II framework in general is flexible enough to say, ‘We’re not going to define a best-execution policy for you, but we are going to demand that you have one,’” he says. 

“‘We are also going to demand that you have an audit trail and that you are able to demonstrate adherence to your best-execution policy.’ This is, generally speaking, in the public good because it implements a regulatory framework to bolster public trust and support the idea of an efficient and fair market,” Adam says.

Algo Trading Post Mifid II

The consensus among market participants, therefore, seems to be that market transparency will ultimately benefit the investor and should make Europe a better place to do business. The regulations will take some getting used to, and the delay to the implementation deadline of Mifid II, once confirmed, will be warmly received. 

That said, when Mifid II does finally arrive, real questions are set to be asked about how financial firms continue to remain both profitable and compliant under the new regulatory regime. The fears that the new regulations have been influenced by mainstream media hype are unfounded, but nevertheless challenges remain. 

As an expert on Mifid II, Meredith Gibson, head of legal risk at Santander UK, says the true test will come when, in the market, questions of profitability are asked. “If you’re talking about best and worst case scenarios, in the best case, it all works like a very well-oiled machine. But in the worst case, it’s a rusty, old Robin Reliant.” The priority for Gibson is not just about compliance, but rather remaining competitive after compliance. “That’s why the big players will just keep on doing what they’re doing,” she says. Gibson says she is positive that post-Mifid II, the industry will work better as a whole than it does now, adding: “We hope.”

Salient Points

•  Mifid II will present the first regulatory framework designed specifically to regulate and monitor algorithmic trading functions within the EU.
• Timestamping, testing and best-execution requirements will all feature as part of the Mifid II regulatory regime.
• Multi-venue testing will curtail the ability for capital markets firms to tweak their algorithms on an intra-day basis.
•  Public mistrust of algorithmic trading firms should soften with a regulatory framework in place supporting the idea of an efficient and fair market.