Hang on a second!

MiFID II introduces the requirement to synchronise the business clocks of trading venues and their customers, standardising the recorded time on post-trade data, transaction reporting and, most importantly, order event auditing. Regulators argue that in the event of unusual market activity they will be able to pinpoint the exact moment things turned sour. A noble cause then, but is it feasible?

The first problem is that of a clock source. Trading systems will need to synchronise themselves to some ‘golden source’ of time. In the interests of avoiding monopolistic ‘Time Lords’ who have legislative authority to charge lots of money for telling the time, it’s probably best to settle on an existing system like GPS. It’s accurate to tens of nanoseconds, but can we use it?

Specialised Precision Time Protocol (PTP) hardware can synchronise clocks in a network at the nanosecond resolution but these are prohibitively expensive – especially for smaller, independent investors running their own hardware. The software implementations of PTP only achieve accuracy to the order of microseconds – that’s thousands of times less accurate than ESMA’s discussion paper speculates will be required for HFT firms.

Assuming you have spent a lot of money on buying a GPS time source, upgrading your network to use it and modifying software to take advantage of the higher resolution clocks, can you meet the requirements? Possibly. Reading the time through off-the-shelf operating systems takes some nanoseconds, writing an auditable event takes some nanoseconds, sending a signal down copper wire takes some nanoseconds. Along with other unpredictable jitters in the system, all of these will conspire to degrade accuracy beyond the acceptable regulated allowance. Perhaps regulating to a few milliseconds’ accuracy may be more achievable?

Regulation is only as good as its enforceability. Given the inherent instability of time measurements at the speculated resolution, how is anyone ever going to provably measure that it is not correct? Are we all just wasting our time?

6 Responses to “Hang on a second!”
  1. I support your comments

    time is money and changes to time will incur expenses

  2. Juan Luis Vicente says:

    Hello, I think systems time synchronization is currently an utopia. I was involved in a similar project and we achieved a good approximation but it never was accurated, the problem was always the same, all data communication has a latency. And I agree with you, these projects are very expensive.

  3. Philip Beevers says:

    The linked paper only talks in terms of microseconds (timestamps should be to the microsecond; everyone should synchronize to the microsecond), and this isn’t all that impractical. As the paper indicates, everybody’s already using GPS as a master clock anyway.

    It’s worth remembering that it’s not actually the time which matters; it’s the temporal ordering of events. It just so happens that coordinated timestamps of appropriate granularity usually provide temporal ordering as a nice side effect. However, in the limit, timestamping alone isn’t enough; the sequence of events needs to be retained too. Also, for some types of events (e.g. any systemic action in response to market data), the regulations have to allow for propagation delays, so again it’s the ordering of events in the receiving system which matters, not the absolute time at which the event was generated at the sender.

    • Robert Turner says:

      Page 518 (14.i) introduces the concept of nanosecond granularity for HFT firms. I think this is what Anne’s referring to.

      • Philip Beevers says:

        I still think the blog exaggerates somewhat.

        The text of the HFT section of the paper reads, “a format that provides granularity of a nanosecond” – they’re careful to imply precision of the timestamp format, not accuracy of the stored data. This looks like future-proofing or even just poor editing given the text elsewhere.

        The word “accuracy” is used in the blog, in the context of clock synchronisation, which is an incorrect interpretation of the material. Section 8.5 of the paper – “Clock Synchronisation” – actually talks about microsecond-accurate clock synchronisation. There’s no implication that nanosecond-accurate synchronisation is needed or even practically possible.

        • Anne says:

          Thanks Phil, good points and I agree it’s not totally clear at this stage in the consultation what will be required. My reading of p522 13 iii) is that the “reportable events” requiring time stamps synchronised to the clock would include orders placed or submitted and stored for NCA’s to request. So there is a potential overlap.

Leave A Comment

Copyright © 2017 Fidessa group plc. All rights reserved.

The information contained within this website is provided for informational purposes only. Fidessa will use reasonable care to ensure that information is accurate at the time it is made available, and for the duration that it remains on the site. The information may be changed by Fidessa at any time without notice. We also reserve the right to close the website at any time. No representation or warranty, expressed or implied, is given on behalf of Fidessa or any of its respective directors, employees, agents, or advisers as to the accuracy or completeness of the information or opinions contained herein or its suitability for any purpose and, save in the case of fraud, all liability for direct, indirect, special, consequential or other loss or damages of whatever kind that may arise from use of the website is hereby excluded to the fullest extent permitted by law. Any decisions you make based on the information in this website are your sole responsibility and information on the website should not be relied upon in connection with any investment decision.

The copyright of this website belongs to Fidessa. All other intellectual property rights are reserved.

Reproduction or redistribution of this information is prohibited except with written permission from Fidessa.