You have been granted access to this page through First Click Free. Subsequent use of TabbFORUM will require logging in. If you don't have an account, registration is free.

Videos

  • Rail_thumb_screen_shot_2012-11-07_at_4

    Threading the Market Data Needle

    Frank Piasecki of ACTIV Financial delves into the difficulties market participants face in extricating themselves from costly enterprise-wide legacy data management systems in favor of faster, cheaper ...
     
  • Rail_thumb_screen_shot_2012-10-15_at_12

    Risk Analytics and Pre-Trade Pricing in Fixed Income

    TABB's Robert Iati and Will Rhode discuss the new partnership between BlackRock and Thomson Reuters to create and distribute a set of fixed income derived analytics to support risk management of fixed income ...
     
  • Rail_thumb_screen_shot_2012-10-01_at_9

    The End of European HFTs?

    Larry Tabb, TABB Group CEO and founder, discusses the findings from the European MiFID II rulings on Equity Market Structure with Adam Sussman, Partner and Director of Research of TABB Group. ...
     
 

More Video | Podcasts

Advertisement
Missing
Graeme Burnett

Morgan Stanley

More From
Graeme Burnett

Spotlight-black
High-Speed Data and OTC Markets

16 November 2012

Putting a Collar on HFT Algorithms

An algorithm can trade 200,000 times before a trader can even react, which poses the conundrum of how these algorithms can possibly be controlled, other than by another algorithm, says Morgan Stanley's Graeme Burnett.

Tick-to-trade time can feasibly be sub 5 microseconds, and for FPGA-based trading algorithms, sub 1 microsecond is achievable for simplistic strategies.

Human reaction time is around 200 milliseconds from visual stimulus to physical reaction, so theoretically at least, an algorithm could trade 200,000 times before a trader could start to react, which poses the conundrum of how these algorithms can possibly be controlled other than by another algorithm?

Perhaps it's time for exchanges to introduce a restricted delta, or "collar," on trading in order to bring volatility back within the domain of human control: for example, the price cannot change by more than 10% in a sliding one-minute period -- giving time for humans to react.

This does not stop high frequency trading -- it just means the algorithms have to accurately follow the "collar" if they want to get filled.

Spotlight-white-trans For more stories in the High-Speed Data and OTC Markets Spotlight Series click here.

Comments | Post a Comment

2 Comments to "Putting a Collar on HFT Algorithms":
  • Missing
    pskopp

    16 November 2012

    Graeme,

    While I am not a fan of the particulars of this plan (http://www.sec.gov/rules/sro/nms/2012/34-67091.pdf) which I believe is overly complex, it is rolling out next year.

  • Anon_avatar
    Anonymous

    20 November 2012

    I've been cogitating on this for several days and I agree that it's over complex unecessarily so. I see little difference between an intraday moving average collar and a start of day price collar with fixed daily percentage limit (e.g. 10% per day). Both achieve the same effect of guaranteeing "market stability" but it's not what I would call a free market. 

    Purely from a technical implementation perspective it's achievable with the usual vagaries of price distribution but will still be happy days for the stat arb snipers gaming the system as they always will (and what is wrong with that)

    Perhaps synchronous price distribution and trade cycle in fixed time quanta is the answer...

You must log in to comment.