MiFID II will see an increased emphasis on justifying choices made when executing a transaction. It requires traders to take ‘all sufficient steps’ to ensure best execution for clients on price, cost, speed, likelihood of execution and more. The key difference come January 2018 will be how firms plan, monitor and ultimately prove they have attempted to achieve best execution.
Speaking to The TRADE earlier this year, Rob Boardman, CEO of ITG Europe, explained some of the implications of the increased focus on conflicts of interest under MiFID II are not as well known as other aspects of the regulation like unbundling, market structure or dark caps.
“On the trading side, there is a lot of scrutiny within asset management firms reviewing their compliance around choosing brokers,” Boardman said. “Regulators no longer accept the reasons often used in the past, instead they want solid evidence and proof.”
MiFID II means the buy-side will be pressured to plan, monitor and demonstrate that an effective best execution process is in place. Alongside this, regulatory technical standards have outlined the need for governance and testing for buy-side firms using algorithms. With countless algorithms available worldwide, firms are tasked with sifting through the noise in order to achieve the best result for clients. A repetitive, yet important task for any trader under MiFID II obligations is to select the right broker and algorithm based on a variety of factors.
“Too many algo choices is a problem if a firm cannot efficiently navigate through them,” says Rich McGraw, senior vice president, global multi-asset EMS/OMS sales, FlexTrade. “If a buy-side firm has hundreds of algos and no way to consistently decipher which ones work well for specific order types, best execution could be at risk.
“Without taking systematic steps, ranking brokers with difficult orders against brokers with easy orders will produce too much inaccurate analyses - or noise.”
Traders are required to have a detailed knowledge of the functioning of the algorithms they use and they must evidence this. In this context, it can be difficult to claim a detailed working knowledge of more than a handful of algorithmic trading strategies and there can be a tendency to refine and focus the list of algorithmic providers.
“The concern expressed by many is that if algo lists are curtailed, how can the trader know that his/her shortlist is and continues to be the best?” says Chris Jackson, European head of Liquidnet’s execution and quantitative services (EQS) Group. “What process do they have for continuous evaluation of both the chosen algo set but also any new algos from different providers?”
Broker randomisation tools, better known as algo wheels - a relatively new concept -burst onto the scene promising traders the ability to assess, monitor and justify algo and broker choices to regulators. With around 1,600 unique broker algorithms to choose from, identifying a clear path through the forest becomes a daunting task.
“The vast choice of algorithms, in conjunction with varying order characteristics and market conditions, makes it very difficult to determine what is ‘normal’ versus ‘outlier’ performance on any particular broker algorithm or order,” says Scott Kurland, co-head of workflow technology at ITG.
“In order to establish a best execution policy, you need to first establish a baseline of normal or expected performance of the strategy, based on a statistically significant set of data and metrics. Without that, every trade starts looking like an outlier.”
Performance driven trading
The overriding concept has been described as being smart routing systems which look at all the options for trading a given asset, and ensure best execution and prevent trader bias.
Trader bias may have previously skewed broker performance, but the algo wheel allows for a larger selection of brokers and even provides a loop of feedback to participating brokers. The underlying basis for the algo wheel is a principal known as performance driven trading.
“Performance driven trading is about reducing the “noise” and trader bias by focusing on a reduced set of variables, such as trading strategy, timeliness and risk tolerance,” says Guy Warren, CEO at ITRS Group. “By using your algo wheel, or randomisation engine, you are truly selecting brokers based on performance metrics, and therefore “best in class” based on your trading parameters.”
FlexTrade takes this a step further and adds intense analysis to the mix. McGraw explains this is at the foundation of FlexTrade’s recently launched algo wheel.
“We like to use the term data-driven trading, which includes both performance-driven trading plus dynamic analyses. Historical transaction cost analysis (TCA) is at the root of both since it can reveal how broker algos performed versus orders with any characteristics,” he adds.
Firms can create trading scenarios such as “VWAP on small-cap names with 20% ADV”, or “Implementation Shortfall on mid-cap names with 10% ADV” or any other conditions. Then TCA can rank how all broker algos performed under similar scenarios,” he explains.
Once this mapping is in place, a trader can randomly select algos with appropriate execution goals and track performance against benchmarks, a step in the right direction for proving best execution. Performance can then be improved through predictive analytics like short-term forecasts or expected volumes, to tweak limit prices and sizes on child orders.
But it’s not just best execution requirements the tool can be utilised for. Algo wheels can be deployed for the unbundling of execution and research, allowing asset managers to effectively assess and demonstrate their approach to broker selection.
The algo wheel can help towards achieving an impartial and best execution objective in the context of MiFID II. With dealers now being asked to do more with less and contribute to the investment process, automating certain parts of the trading blotter can similarly make them more efficient.
However, there is a danger moving away from manual processes could have a negative impact on trading performance.
“Algo wheels can work relatively well to make the trading of the ‘low touch’ portion of the blotter more efficient but we have seen examples where over-automation of the execution process around more complex trades has resulted in significant execution missteps,” says Jackson.
“There is a danger that over-automation is a first step to commoditisation of the trading process and risks dumbing down the skilled and nuanced process that is trading.”
He adds there have been examples of situations where algo wheels and other multi-broker allocation processes have been either badly managed or subject to significant conflicts of interest occasionally resulting in manipulation of the allocation logic to favour one party over another.
For an algo wheel to achieve the best results the provider must be impartial.. This means that for a broker dealer operating their own algo - if also providing a wheel - should not be an algo destination on that wheel. Secondly, the provider must have a transparent and independent mechanism when used for assessments of transactions. And lastly, it should include continuous and impartial assessments of new alternative products to ensure wheel participants are best in-class.
With ITG and FlexTrade already off-the-mark – at the time of publication - with the launch of their respective algo wheels, we can expect other financial services firms to follow suit as MiFID II’s best execution requirements take centre stage in January 2018.