By Scott Kurland and Daniel Shaw, Managing Directors at ITG
Depending on one’s perspective, buy-side traders today are either spoiled for choice or overwhelmed by options. At last count there were more than 1600 discrete algorithms available worldwide. The unique parameters (and sometimes quirky names) associated with each of these algos makes selection, performance measurement, and broker feedback more difficult than ever. Combine this with rapidly changing market conditions, varying order characteristics, liquidity constraints, and trader preferences, and it becomes clear that institutional investment managers face a near-impossible task of demonstrating a coherent best-execution policy.
With MiFiD II looming, the buyside is under heightened pressure to demonstrate that they have a best-execution practice that they can explain and defend. In addition, Rule 6 of the MiFiD II Regulatory Technical Standards (RTS 6) will impose governance and testing obligations for all investment firms running algorithms, which will put more of a burden on the managers themselves in the form of time and resources needed to test each new broker algorithm suite and parameter set. Investment managers will also need to demonstrate that they have a global “kill switch” for all open/working orders. RTS 6 will potentially impact any EMEA-based investment manager, as well as any global investment manager who has client assets/accounts, or a local desk presence in the UK or Europe.
Much fanfare has been made lately by some OMS, EMS and agency broker vendors who claim to solve this problem by designing so-called “intelligent switching” and routing engines that leverage machine learning to automatically select and continuously modify the optimal broker, algo, and strategy for any given order at any given time. While this may sound alluring, this approach further complicates the process, compounding the challenges at hand.
The Pitfalls of “Intelligent Switching”
- Less Insight – The buy-side trader gains even less insight into what and why the engine is choosing or how it is switching its selection.
- Misused Algos – The constant switching (cancel/correction) of orders midstream does not permit a receiving broker’s algorithm to complete its objective based on the original order inputs – thus, algos are being used incorrectly.
- Higher Maintenance – The need to refresh algo specs, update broker screens and map FIX tags results in frequent updates to desktop EMS/OMS software and destination hot-buttons.
- Increased Costs - The switching can also increase ticket-splitting, resulting in higher back-end settlement costs associated with processing more trades.
- Muddled Measurement – Measuring relative performance across the broker universe becomes more challenging, absent a statistically significant set of ”apples-to-apples” data. This noisy data can also impact the accuracy and utility of transaction cost analysis (TCA).
- Disrupted Feedback Loop – The misuse of algos by the switching engine, coupled with ambiguous measurement metrics, denies brokers the accurate feedback they need in order to retool their algorithms and improve performance.
These headaches are borne not only by the investment manager, but are also felt by their brokers, their EMS/OMS/FIX vendors and their TCA providers. In short, the only party truly benefiting from an “intelligent switching” engine is the vendor providing it.
Given these challenges, a few innovative institutional managers have come up with an alternative solution. It’s called Performance Driven Trading, and at ITG we have had the privilege of collaborating with these managers to bring this alternative solution to the market.
“Precision measurement of broker algo performance has always been a challenge. The Triton Algo Wheel that Schroders has developed in conjunction with ITG provides a consistent global approach to quantitative performance evaluation within our broader best execution process,” states Gregg Dalley, Global Head of Equity Trading at Schroders in London.
The Algo Wheel is a key feature in Performance Driven Trading. This approach seeks to simplify the process, reduce the potential for trader bias, and provide an evolutionary path towards relative performance measurement with continuous process improvement, in collaboration with the broker community.
The Performance Driven Trading Approach
- The first step involves reducing the variables in the decision and selection process, recognizing that the core inputs for any trading decision are primarily the manager’s trading objective (or benchmark), time horizon, and risk tolerance. These factors drive the strategy selection (i.e. VWAP, IS, dark, etc.).
- Once a strategy has been selected, only a common, critical set of parameters related to the core inputs are set (such as start time, end time, urgency level, and participation rate).
- To reduce or eliminate trader bias, broker selection is left to a randomization engine, or “algo wheel” – subject to constraints or goals (broker restrictions, participation limits, etc.).
- Because the strategy set and core inputs have been normalized, a broader universe of brokers can be included. The process allows for a statistically significant, unbiased data set to be gathered across multiple orders and varying market conditions.
- This yields an opportunity to develop accurate broker rankings, with ”apples-to-apples” context and much of the noise removed, based on proper use of each algorithm on any given order. Brokers then have the ability to truly benchmark their performance versus peers and to re-tune their algos to improve performance.
- In order to reduce information leakage and ticket-splitting, the solution can be configured to send all follow-on orders in a given security and trading day to the broker who received the original order.
- Ultimately there is room for machine learning to be applied here – but it should be used to help optimize the manager’s three core inputs: trading objective (strategy selection), time-horizon (urgency), and risk tolerance (participation levels).
Performance Driven Trading and MiFiD II
The Performance Driven Trading approach may also help simplify the self-certification requirements imposed by RTS 6 by streamlining and reducing the number of broker and algorithmic strategies, which need to be validated, and distilling this into a normalized set of common, critical parameters. Performance Driven Trading is actually a complete process and not just a technology solution; it provides the data required for broker reviews and relevant performance feedback, the opportunity to identify and quantify performance improvement across brokers over time, and the ability to define, explain and justify a best execution practice to both clients and regulators. Given the increased administrative burdens of MiFiD II and the heightened competition in the asset management industry, implementing a Performance Driven Trading solution is a true intelligent switch.