RegTech Intelligence

Know thy algo: how to define it, prove it, tame it. Part 2

JWG analysis.

The desire for tighter controls on algorithmic trading is growing globally.  Trading rules in the major financial centres will quickly set new minimum thresholds.

As described in our previous piece on how algos are defined and controlled, Europe is again leading the pack and would appear to have serious intent to change the way algorithmic trading is supervised.  The results will have wide ranging impacts on the commercial nature of HFT.

Question 4: how will strategies be monitored?

National competent authorities will be empowered, under MiFID II, to request regular ad-hoc descriptions of firms’ algo strategies for monitoring purposes.  ESMA have made it clear that, in order for a firm to comply with this requirement, they must keep records of any material changes made to their proprietary software, allowing them to accurately determine when a change was made, who made the change, who approved the change and the nature of that change.

Clearly this will create a large amount of new papering requirements for firms, but what we do not yet know is whether the authorities will actually make use of these new powers.  The more critical question, however, is how will firms disclose commercially sensitive information that could destroy their business?  Surely, the whole point of rewarding traders that develop better mousetraps with intellectual property rights is under attack.  Answers to question 205-206 of the discussion paper etched into those stone tablets, quickly…

Question 5: what should be in the reports?

As stipulated by article 26 of MiFIR, it will become necessary to flag all algo generated orders in transaction reports so that regulators will be able to reconstruct firms’ strategies.  Here, the Europeans appear to be taking their lead from BaFin, although how exactly this requirement will line up with the German HFT Act is, as yet, left unaddressed.

Firms will have responsibility for identifying their own algos, provided that one is distinguishable from another.  The method of determining which algos are distinct from each other will be crucial to this rule.  And whether this will be sufficient for regulators to identify algos is also questionable.

Either way firms will have to figure out a way to put the processes in place in order to be able to do this.  This raises issues such as how to put in place a set of identifiers for all the moving parts of a trading system, and how to be able to find and prove that the process running on a virtual machine that is no longer there was in, fact, correctly identified.  Answers to questions 552-554 from all those metaphysical magicians out there, please.

Question 6: what should be recorded?

MiFIR article 25 sets out that firms must keep records of all orders, including the order ID, time and date and characteristics.  In their discussion paper, ESMA made it clear that this is aimed squarely at HFT, further elaborating that records must be time sequenced and include cancellations, executed orders and quotes, all of which must be available to competent authorities on request.

These requirements are designed to give regulators greater oversight and monitoring power over algorithmic trading, with ESMA again taking a leaf out of BaFin’s book.  But in order for this to work in practice, a clear definition of what constitutes an order is necessary, especially in the context of HFT.  Obviously, firms’ record keeping policies will need to be drastically overhauled.  If you have views on this issue, look to questions 260 and 261 in the discussion paper.


What impact will all of this have on the market?  Current thoughts range from little to devastating but, ultimately only time will tell.  What is clear is that the industry needs to focus on implementing the systems and workflows that will be needed in order to meet all these new requirements – and that won’t be easy.

As noted in part one of this article, this is just a starter set.  MiFID II is a mass of interrelated issues and decisions made in other areas of the consultation and discussion papers will undoubtedly affect the questions we have scoped out here.  In a nutshell, we are going to be struggling to really know our algos for some time to come.

At the end of the day, we will have more certainty with the technical standards by Q2 2015.  And implementation will be a mere 12 months later.  The industry will need to be ready to move its infrastructure in a relatively short window.

We will be having a robust debate on ‘what good looks like’ at our next CDMG meeting.  Get in touch if your firm isn’t amongst the 15 already registered.

To promote global dialogue on how to deliver regulatory change JWG post hundreds of focused articles a year to thousands of subscribers. Get involved and join the mail list.

By hitting the subscribe button you agree to our Privacy Policy