The Concept of ‘Throttling’ High-Frequency Trades
|Sep 10th, 2013 | Filed under: Algorithmic and high-frequency trading, Today's Post | By: cfaille||
The Commodity Futures Trading Commission has published its much-anticipated concept release on high-frequency trading and related issues.
Traditional risk controls, the CFTC observes, were designed for human judgment and human speed, and they must be reevaluated in light of the “electronic trading platforms that can execute repetitive tasks at speeds orders of magnitude greater than any human equivalent.”
To the agency’s credit, the report is thoroughly researched and carefully organized. It includes 124 specific questions addressed to commenters. Indeed, so thorough is it that a blog entry like this, at our customary length, can do little more than pluck our some of the more fascinating chunks. I will note, though, that my earlier speculations about what the report wouldn’t say have turned out to be right.
Public comments on the release are welcome within a 90-day period.
The Question of Definition
In some respects the CFTC is trying to build on an earlier report by a working group of its Technology Advisory Committee, which proposed a definition of HFT incorporating (a) algorithms that execute individual transactions without human direction; (b) low-latency technology including proximity and co-location; (c) high-speed connections to markets for order entry; (d) high message rates (orders, quotes, and cancellations).
As the TAC working group observed, there are various ways in which one might measure how high is high in terms of those message rates. There are cancel-to-fill ratios, participant-to-market message ratios, or participant-to-market trade volume ratios.
The CFTC now wants commenters to offer it guidance on the following points:
- In any forthcoming rule-making, should the CFTC adopt a formal definition of HFT?
- Should it adopt the TAC definition? With or without amendments?
- Are the three objective measures of recurring high message rates stipulated in last year’s TAC report sufficient to distinguish between automated trading systems in general and those that use HFT strategies? For each of those three metrics, what is the pertinent threshold value?
- Should the definition of HFT (whatever it turns out to be) make any regulatory difference? That is, should “the risk controls for systems and firms that engage in HFT be different from those that apply to ATSs in general?”
With an eye to risk management, the CFTC inquires what the industry thinks of increasing the standardization of real-time order, trade, and position reports, by market participants and clearing firms alike.
The Issue of a Throttle
In this context, it cites a report by Carol Clark and John McPartland, prepared for the Chicago Fed in May 2012, which detailed the current risk-management practices of nine proprietary trading firms. Clark and McPartland found that all nine firms have maximum order sizes and intraday position limits, all have kill buttons, and eight have credit limits by account, in order to monitor open positions, the dollar value of those positions, and the quantity of working orders. The CFTC wants comments on “what types of risk controls are most commonly used throughout the industry, and the degree to which those risk controls are standardized throughout the industry.”
Another subject on which it desires comment is the aptly named “execution throttle.” The term “throttle” is a borrow from mechanical engineering, where it refers to a lever, pedal, etc. used to control a value, and thus to slow or stop the flow of a fluid (typically fuel or air or both).
In the CFTC’s borrowing, an execution throttle is a system that determines when an algorithm has exceeded a given message rate or rate of execution. When tripped, the throttle alerts monitors at both the exchange and the trading firm.
Specific questions along these lines include:
- “If, as contemplated above, maximum message rates and execution throttles were used as a mechanism to prevent individual entities or accounts from trading at speeds that are misaligned with their risk management capabilities, how should this message rate be determined?”
- Also, if throttles are to be implemented at the trading firm level, the CFTC asks whether they should be “applied to all ATSs, only ATSs employing HFT strategies, or both?”
So, again, the latter of those questions shows that it remains an open question not just how to define an HFT but whether it matters how one defines it.
Two of the members of the CFTC, Bart Chilton and Scott D. O’Malia, added “concurrences” as appendices to the concept paper. Though not dissents, the concurrences did manage to put a little distance between each of these two and the paper proper.
Chilton expressed his frustration with the “puny penalty regime at the CFTC,” which can only attach civil monetary penalties of up to $140,000 per violation. With that petty deterrent, it is clear he said that whatever the rules are or will be, traders will continue breaking them in the face of vast potential profits.
O’Malia expressed regret that the report did not offer a “more thorough and clear cataloguing of existing industry practices and recommendation,” but adds that a recent TAC reference document does a better job of this.
Christopher Faille is a Jamesian pragmatist. William James has taught him, for example, that "you can say of a line that it runs east, or you can say that it runs west, and the line per se accepts both descriptions without rebelling at the inconsistency."