By continuing to use this site you are agreeing to our use of cookies, as set out in our privacy policy.

« go back Comment MIFIDII_Algos.jpg

MIFID II for algos – regulatory validation and model management

The automation of trading decisions has led to lower costs and greater liquidity but has the potential to amplify execution risks, turning otherwise manageable errors into extreme events. The MIFID II articles on Algorithmic trading aim to mitigate this risk and ensure a level of independent oversight. We suggest a robust industrialised approach for managing Algorithmic trading models, bringing them in-line with best practice and taking into account some of the unique challenges in this area.

Colin Dick 26 November 2018

The automation of trading and execution decisions has led to lower overall costs and greater liquidity in flow capital markets, but, as the FCA’s recent paper points out, algorithmic trading has the potential to ‘amplify’ certain risks, turning ‘otherwise manageable errors into extreme events with potentially wide-spread implications’.

Implementation of the MIFID II articles on Algorithmic trading helps mitigate this risk of extreme events in so far as conforming to the articles forces best model management practice on algorithmic trading teams and ensures a level of independent oversight.

Based on our experience in traditional front- and middle-office model validation for very broadly, pricing and credit and market risk, Riskcare has advocated a lifecycle approach to ‘industrialised model management’ (IMM) that aligns to, e.g., SR 11-7 on model risk, as contrasted to more piece-meal, traditional ‘model validation’ approaches.

This approach breaks down the lifecycle into design/implementation and execution phases, defines an operating model ensuring strict independence of ownership, test and development and promotes iterative improvements.

Clearly, one common theme for the regulators is that independent validation and management of strategies is key to effective risk management.

For algorithmic trading that means independence from the front-office that builds the strategies. IMM would appear to be a perfect fit for bringing Algorithmic trading models in-line with best-practices and satisfying MIFID II regulations on independent validation and management.

However, our conversations and experiences with banks around this subject have highlighted areas where the nature of algorithmic trading and the detail in MIFID II makes moving toward the ‘perfect’ independent approach hard.

This paper briefly outlines Riskcare’s optimal model management vision and discusses some of the challenges in aligning algorithmic trading with that vision to provide a truly independent approach.

 

Vision for industrialised model management

The IMM approach divides model management into Design, Implementation and Runtime. For each of these stages we recommend features of best-practice to be adhered to as well as advocating that the three stages be looked at holistically.

The proposal is that a bank adopting IMM as a vision for model management will look to apply this standard governance approach across all front- and middle-office models and that models will progressively move to a central framework to support IMM.

Table.jpg

The operating model around the IMM framework advocates independent ownership of lifecycle events, so that design and implementation teams are independent from validation and testing teams in accordance with SR 11-7. We also recommend that senior managers have overall ownership of the models and have enough consolidated information to comfortably sign-off changes.

The approach to working within the framework is iterative, so that feedback from runtime experience with models informs changes to design which feeds through to implementation and runtime.

Thus stated, the IMM approach aligns to a modern dev-ops approach to the model life cycle. The key business benefits from standardising the model development process are:

  • Quicker and safer turn-around of models in production.
  • Strict independence of validation from development and ownership.
  • Consistency in outputs, regulatory reporting, regression testing.
  • Conformity to regulatory standards such as SR 11-7 and BCBS239 and consistency in proving conformance.
  • Model ‘cohesiveness’, for example, present-value models for credit risk PFE and front-office pricing are managed similarly, even if they need to use different implementations for performance reasons.
  • Making model dependencies explicit in the inventory supports independent unit-testing and validation of the component.
  • Cost-savings in human and physical resources through standardisation.

The IMM vision provides an appealing strategic state that can be become a blueprint to guide banks as they seek to improve their infrastructure and operating models. However, in practice, when building and managing classes of financial models within the IMM vision, certain challenges will almost certainly be encountered. In what follows we examine the MIFID II algorithmic trading requirements and see how far it is practicable to satisfy them by moving towards the IMM vision...

To view the full article, please click the link below.

Open the MIFIDII for algos PDF