Algorithmic Trading Briefing Note

1y ago
940.54 KB
8 Pages
Last View : 4d ago
Last Download : 5m ago
Upload by : Adele Mcdaniel

Senior Supervisors GroupAlgorithmic TradingBriefing NoteLLO F T HEERCRENCYURCOMPTROApril 2015186 3

Algorithmic Trading Briefing NoteTABLE OF CONTENTSI.Introduction and Market Evolution. 1II.Key Risks.2III. Key Control Principles and Sound Practices.3IV. Next Steps: Questions for Firms and Supervisors.4V.Conclusion.6

Algorithmic Trading Briefing NoteExecutive SummaryHigh-frequency trading (“HFT”), or high-speed trading1(“HST”), a type of algorithmic (or “algo”) trading, is nowa well-known feature of the global market landscape. Inmany markets, a small number of firms may account fora large proportion of trading volume. Although it hasbeen argued that HFT has lowered investors’ trading costsby reducing bid-ask spreads, the risk that HFT activityspecifically, and algorithmic trading more generally, posesto firms and the financial markets has sparked debate andraised concern among market participants and regulatoryagencies globally. This is, in part, owing to the speedof trading and, therefore, the pace at which exposuresmay accumulate intraday at financial institutions. Indeed,unexpected events linked to algorithmic and highfrequency trading have caused significant volatility andmarket disruption, leading to heightened debate aroundthe risks these activities pose to the functioning of globalmarkets. The complexity of market interactions among HFTfirms and other market participants increases the potentialfor systemic risk to propagate across venues and assetclasses over very short periods of time.This briefing note focuses on how risks associated withalgorithmic trading are monitored and controlled at largefinancial institutions during the trading day. While marketstructure and trading rules differ by jurisdiction and assetclass, we seek to identify risks common to algorithmictrading and to suggest questions that supervisors might1For purposes of this paper, the terms “high-speed trading” and“high-frequency trading” refer to automated trading conducted atmillisecond or microsecond speeds throughout the trading day.The Senior Supervisors Group (SSG) is composed of the staff of supervisoryagencies from ten countries and the European Union: the CanadianOffice of the Superintendent of Financial Institutions, the EuropeanCentral Bank Banking Supervision, the French Prudential Controland Resolution Authority, the German Federal Financial SupervisoryAuthority, the Bank of Italy, the Japanese Financial Services Agency,the Netherlands Bank, the Bank of Spain, the Swiss Financial MarketSupervisory Authority, the United Kingdom’s Prudential RegulatoryAuthority, and, in the United States, the Office of the Comptroller ofthe Currency, the Securities and Exchange Commission, and theFederal Reserve.The SSG acknowledges the important contributions of Caleb Roepe,George Wyville, and Stephanie Losi to this briefing note.consider as they monitor or examine this activity. Further,by setting forth risk-based principles and questions thatfirms already engaged in algorithmic trading can use toassess their controls over this activity, we aim to facilitatean informed conversation about sound risk managementpractices and renew market participants’ focus onimproving risk management of this activity.Key supervisory concerns center on whether the risksassociated with algorithmic trading have outpaced controlimprovements. The extent to which algorithmic tradingactivity, including HFT, is adequately captured in banks’risk management frameworks, and whether standard riskmanagement tools are effective for monitoring the risksassociated with this activity, are areas of inquiry that allsupervisors need to explore. Further, algorithmic tradingactivity has expanded beyond the U.S. equity markets toother markets and asset classes, including futures, foreignexchange, and fixed-income markets. Thus, our supervisoryapproach needs to remain flexible and adaptable to addressgrowth and evolution of this activity.I. Introduction and Market EvolutionAlgorithmic trading has grown and evolved in responseto the many changes that have taken place in the marketlandscape since electronic communications networks(ECNs) became established in the late 1980s into the 1990s.Advancements in trading technology, along with regulatorydevelopments, have played a role in fundamentallychanging the structure of markets and the way thatsecurities and derivatives are traded. HFT activity, forexample, benefited from technology that reduced delay,or latency, to the markets. In addition, by co-locatingtheir servers with market servers at an exchange or darkpool data center, algorithmic traders, including HFTfirms, increased the speed at which they could accessthe markets. The decimalization of pricing helped inthe advancement and development of certain forms ofalgorithmic trading, such as HFT. Meanwhile, alternativetrading systems (ATS) and dark pools began to attractvolume and grow market share. Changes in the marketlandscape, such as growth in the number of electronicallytraded markets, also played a role in the expansion of,1

Algorithmic Trading Briefing NoteModel of U.S. Equity Market rDarkpoolBrokerdealerAlthough certain types of algorithmic trading may reduceperceived bid-ask spreads, algorithmic trading alsoincreases operational risk at individual firms and acrossthe financial system. For example, an algorithmic strategymay fail to maintain risk exposure within a specifiedthreshold during volatile market conditions, or fail toreturn to an allowable risk position. An undetected failurewith one market participant’s algorithmic trading strategycan increase, or further transmit, risk to another firm, orto the markets more generally. As algorithms and theirinteractions grow in both number and complexity, varioustypes of algorithmic trading may increase systemic risk.BrokerdealerII.and investments in, algorithmic trading. The combinationof these and other factors facilitated the overall growthin algorithmic trading; however, no single factor explainsthe growth and evolution of the different subsets ofalgorithmic trading.What is quite clear, however, is that algorithmic tradingis a pervasive feature of markets in many countries, eachwith its own regulatory structure. Often associated with theequities market, HFT, for example, spans asset classes andtraded products across jurisdictions and trading venues.HFT in foreign exchange (“FX”) and rates markets, whichhave different regulatory controls than equities markets, hasgrown substantially over the past decade. Additionally, firmsengaging in algorithmic trading activity have benefitedfrom a fragmented market structure in at least one way—by arbitraging prices between different trading venues.Many traders’ roles also changed as algorithmic tradingevolved and expanded. Instead of making order executiondecisions based on valuation models or in the course ofmaking a market or facilitating clients’ orders, traders nowuse trading strategies based on algorithms to arbitrageprice differences across related products and tradingvenues and take advantage of liquidity, or lack thereof, indifferent markets. Traders can adjust an algorithm withincertain boundaries (for example, tuning an algorithm to trademore or less aggressively), but actual orders are generatedby the algorithm based on response to market signals.For instance, many banks employ algorithms designed toexecute trades without significantly impacting market prices.Key RisksRisks associated with algorithmic trading, including HFT,have been covered extensively in numerous forums, suchas the International Organization of Securities Commissions(IOSCO) consultative report2 and market regulatorroundtables3 and concept releases4. We will not repeat thediscussion of the risks already covered in these reports andforums. Instead, we take the opportunity in this briefingnote to lay out four risks commonly associated withalgorithmic trading:1. Systemic risk may be amplified. An error ata relatively small algorithmic trading firm maycascade throughout the market, resulting in asizable impact on the financial markets throughdirect errors or the reactions of other algorithmsto the error. Clearinghouses and centralcounterparties (CCPs) may also be affectedby erroneous trades, though their degree ofexposure to clearing members may be limiteddepending upon the product category involvedand the nature of the alregister090913.pdf5In addition to numerous procedural safeguards and financial resourcesdesigned to anticipate and absorb extreme but plausible exposures,clearinghouses and CCPs have in place, or are developing, safeguardsspecifically designed to control exposures that may arise as a result oferroneous trades.32

Algorithmic Trading Briefing Note2. Algorithmic trading desks may face asignificant amount of risk intraday withouttransparency and robust controls. Proprietaryand agency trading desks at financial institutionsgenerally have risk reporting and risk controls inplace to limit and control risk exposure acquiredover the course of the trading day. However, atany institution, intraday risk controls may not berobust, reporting may not be complete or timely,or limit breaches may not be transparent to seniorrisk officers. As such, an unintended accumulationof a large position during the trading day mayresult in a firm taking on significant exposurebefore end-of-day risk processes take effect. Also,technology failures, exceptional or unanticipatedmarket conditions, or an unexpected failure ofan algorithm during the day could force a firmto carry significantly more risk overnight thanit had intended, and without timely senior riskmanagement oversight.3. Internal controls may not have kept pace withspeed and market complexity. Malfunctionsand outages at financial institutions and criticalentities such as exchanges are not new, buttheir potential impact can be amplified. Further,many banks’ prime brokerage businesses havealgorithmic and HFT firms as clients, and therisk controls and monitoring efforts acrossthese businesses vary widely among banksand are evolving to keep pace with this typeof client activity.4. Without adequate controls, losses canaccumulate and spread rapidly. Examples ofrapid-impact events include the 2010 Flash Crash(a large-order execution algorithm operating inan unexpected way), the 2012 Facebook IPO(an exchange system problem), and the 2012Knight Capital incident (the malfunction of anorder routing system). As follow-ups to theseevents, published regulatory enforcement actionshighlighted control shortcomings related toinsufficient testing6, and new rules and recentproposals by market regulators have aimed toincrease control 3/34-69655.pdf (NASDAQFacebook IPO) and pdf(Knight -67091.pdf (Limit Up-LimitDown) and df(proposed Regulation SCI).III. Key Control Principlesand Sound PracticesThe following is a list of principles for supervisors toconsider when assessing practices in and key controlsover algorithmic trading activities, including HFT, at banks.These key controls and practices are both preventative anddetective in nature, and are the underpinnings of an overallrisk management framework.Controls must keep pace with technological complexityand trading speeds. A multilayered “defense-in-depth”strategy, which increases control redundancy and diversity,can reduce the risk that an erroneous or destabilizingorder will reach financial markets. Defense-in-depth isa concept from the information security field that callsfor multiple controls at multiple points in a process. Forexample, algorithmic trading controls should exist prior toalgorithm launch or change deployment, during the tradelifecycle, and during the incident response process. Firmsneed to have controls that cover all aspects of the tradingprocess, including order generation, order handling, andorder execution.Governance and management oversight can limitexposure to losses and improve transparency. Onesound practice is to establish firm-wide governance foralgorithmic trading controls being aligned with the firm’sstated risk appetite framework and apply it consistentlywithin the firm. Without such consistently appliedgovernance, differences in controls across desks canintroduce unnecessary risk for a firm and can representa lost opportunity to identify and implement bestpractices across all desks.Testing needs to be conducted during all phases of atrading product’s lifecycle, namely during development,rollout to production, and ongoing maintenance. Initial testing: Firms wishing to deploy a newor updated strategy or algorithm must firstconduct simulations and non-live testing withina trading venue testing environment. Appropriatetesting helps to ensure that algorithms pass therisk management controls required by the firmand the exchange. Controlled rollout: The algorithms shouldbe rolled out in a controlled and cautiousfashion. Initially, a firm should self-imposeprice and position limits as well as limits onthe number of instruments and venues wherethe algorithm is deployed.3

Algorithmic Trading Briefing Note Ongoing testing: Firms should test systemsand controls to ensure that they can withstandsignificant or elevated market volumes andexternal events that could exert stress on thosesystems and controls.When assessing control depth and suitability,management should ensure sufficient involvementof control functions (such as Compliance, Technology,Operations, Legal, Controllers, and Market Risk) aswell as business-unit management. All stakeholdersshould have a voice at the table in determining the rightbalance between risk and controls. Control functions anddevelopers both need to understand the inherent risks thatalgorithmic trading poses to ensure that the proper controlsare in place. Defense-in-depth also gains strength whendifferent layers within a firm play a role.IV. Next Steps: Questions forFirms and SupervisorsIn this briefing note, we address three specific oversightlevels: business unit or desk management; control functionsand senior management; and the board. For each level,we present questions that supervised firms can use toself-assess their current control state and risk appetite.As a corollary, supervisors also may consider askingfirms these or similar questions.Business-Unit/Desk-Management Level hat degree of role overlap is permitted among traders,Wstrategists, and developers?Situations in which front-office traders, strategists, anddevelopers lack clearly defined role boundaries can resultin insufficient oversight and create conflicts of interest.Management should consider the risks of allowing tradersto develop their own algorithms; of allowing developersto test their own code; and/or of allowing developersto deploy their own code into production. All of thesesituations run counter to separation-of-duties principles andmay increase the likelihood of errors. Fast-paced, iterativedevelopment environments require heightened controls.Does the unit adhere to firm-wide policies and processes?If front-office trading desks develop their own processessuch as change management rather than followingfirm-wide processes, risk may exceed the level intendedby senior management and the board. Therefore,independent risk management functions shouldreview and approve any changes and should informsenior management and/or the board of these exceptions.Moreover, independent risk management should revisitdesk-specific procedures periodically. oes the unit seek advice from independent controlDfunctions to ensure that compensating controls areeffective and functioning as intended?If resource or personnel constraints prevent a tradingdesk from implementing optimal controls, the likelihoodof generating erroneous or destabilizing orders increases.The business unit should establish compensatingcontrols to mitigate both firm-specific and systemic risk.Independent control functions at the firm should verify,approve, and periodically reassess the adequacy of thesecompensating controls.What types of risk reports does the unit produce,and who in the firm receives those reports?If business units do not provide transparency about risktaken intraday, senior management and the board mayremain unaware of the business’s full risk profile. Toprovide transparency, trading desk management shoulddevelop appropriate reports (for example, intraday profitand loss) for independent risk management and/or seniormanagement and the board.Control-Function and Senior-Management LevelDoes an independent risk management functioncommunicate with senior management about thelevel of intraday risk taken by each desk at thefirm? Can independent risk management effectivelychallenge the front office if they identify excess risk?In line with the recommendation that desk managementshould develop appropriate risk reports, independentrisk management should review those reports and ensurethat senior management is aware of the level of intradayrisk taken across the firm. In addition, independent riskfunctions should be robust enough to develop their ownkey risk indicator reports and provide them to a chief riskofficer, risk committees, or similar management governancebodies. If risk managers feel that the exposure created bycertain business activities or transactions is too great giventhe firm’s risk appetite, they should have the power to delayor decline activity pending senior management briefing4

Algorithmic Trading Briefing Noteand consideration. In particular, senior managementmust have an understanding of how rapidly positions canaccumulate intraday. Independent risk management and/orsenior management also should engage with and challengebusiness units as needed to ensure that intraday riskexposure remains within acceptable limits.Are control-related functions such as Technology,Operations Management, and Compliance aware ofthe controls in place on trading desks, and do theyview those controls as sufficient? Do they have theability to mandate stronger controls if they perceivegaps or weaknesses?If control functions have only limited input or insight intofront-office controls, firms may miss opportunities to alignwith industry best practices. They may, furthermore, facelosses or liability if a control gap allows an erroneousor destabilizing order to reach financial markets. Robustinvolvement in the control-setting process by controlfunctions can foster productive discussion about theoptimal balance of control strength and innovation pace,resulting in a stronger, more resilient overall business.Furthermore, control functions are key secondary andtertiary lines of defense for firms. Their own controlframeworks can be important supplements andcomplements to those of the front office.Who receives reports on major incidents and/or losses?Is senior management aware of glitches and incidents?How does the firm communicate lessons learned fromany incidents and/or losses?Near-misses that are not reported represent lostopportunities to strengthen controls broadly across afirm. By reporting near-misses through firm-wide incidentmanagement systems, even when no financial loss hasoccurred, firms gain the opportunity to mitig

High-frequency trading (“HFT”), or high-speed trading1 (“HST”), a type of algorithmic (or “algo”) trading, is now a well-known feature of the global market landscape. In many markets, a small number of firms may account for a large proportion of trading volume. Although it has

Related Documents:

Algo trading TOTAL TRADING ALGORITHMIC TRADING HIGH FREQUENCY TRADING . Algorithmic trading: In simple words an algorithmic trading strategy is a step-by-step instruction for trading actions taken by computers (au

Algorithmic Trading Table: Proportions of trading volume contributed by di erent category of algorithmic and non-algorithmic traders in the NSE spot and equity derivatives segment (for the period Jan-Dec 2015) Custodian Proprietary NCNP Total Spot Market Algo 21.34% 13.18% 7.76% 42.28% Non-

Introduction: Algorithmic trading Developing an automated trading decision engine – Identify a successful trading rule – Extend trading rule set – Automate trading rule selection Implementing MATLAB int

the algorithmic trading strategy’s design; typically, broker algorithmic trading systems seek to minimize the cost of trading by optimizing the execution strategy—that is, minimize market impact cost or time to execution, optimize the price, and so on—whereas proprietary algo - rithmic

trading is not inherently harmful to market integrity or quality – in fact algorithmic trading plays an important role in developing liquidity. Equally, algorithmic trading is not, by definition, inherently good for market quality and integrity. Significant order f

Algorithmic trading (automated trading, black-box trading, or simply algo-trading) is the process of using computers programmed to follow a defined set of instructions for placing a trade in order to generate

Algorithmic trading is any software which follow a predefined algorithm to place trading instructions High-frequency trading is algorithmic trading characterized with very

The Effect of Algorithmic Trading on Liquidity in the Options Market Abstract Algorithmic trading consistently reduces the bid-ask spread in options markets, regardless of firm size, option strike price, call or put option, or volatility in the markets. Howeve

Algorithmic Trading System in question (see Good Practice Statement 3 for more detail); sufficiently granular oversight of the financial and non-financial risks relating to Algorithmic Trading (which, depending on the size of the firm, may include establishing regional, product or as

Jan 07, 2016 · to resume trading after an event —there may be scope for firms to wind down their operations instead. Market making strategy MiFID II imposes obligations on algorithmic traders when they pursue a market making strategy. A person engaged in algorithmic trading will be consid

of algorithmic trading over the last two decades. The fi rst is the fact that the fi nan-of algorithmic trading over the last two decades. The fi rst is the fact that the fi nan-cial system is becoming more complex over time, not less. Greater complexity is a consequence of general economic growth and globalization in which the number of

ADVANCED ALGORITHMIC TRADING STRATEGIES IN PYTHON Learn to find alpha in financial markets using artificial intelligence and quantitative techniques. Get trained by experts such as Dr. Ernest P. Chan, with decades of trading experience. Learn to code and backtest your own trading strategy in Python

Any strategy for algorithmic trading requires an identified opportunity that is profitable in terms of improved earnings or cost reduction. In this paper, the study focuses on evaluating few trading strategies and trading . we used a Quantopian platform (Algorithm IDE), which is a Python

Market Microstructure and Algorithmic Trading PIMS Summer School 2016 University of Alberta, Edmonton Lecture 2: July 4, 2016. Edmonton mini-course, July 2016 Exchanges Decide trading rules: Trading hours Match alg

Algorithmic Trading and Information Terrence Hendershott Haas School of Business University of California at Berkeley Ryan Riordan Department of Economics and Business Engineering Karlsruhe Institute of Technology August 18, 2009 Abstract We examine algorithmic trades (AT

the place of anatomy in medical education amee education guide, service manual 1988 mercury 70hp, atlas copco fx 6 manual, force 120 hp outboard repair manual, options trading beginners guide to make money with options trading options trading day trading stock trading stock market trading and investing trading volume 1, gerhard richter, your flight

GearboxNDX indicator and trading rules! ! ! ! !43-47 GearchangerNDX indicator and trading rules! ! ! !48-50 Trend and Stops indicator and trading rules!! ! ! !51-54 Volume indicator and trading rules! ! ! ! ! !55-57 Volume Radar indicator and trading rules! ! ! ! !58-60 No Progress ndicator and trading rule

Trading system 7: Donchian style breakout Trading system 8: Breakout with EMA confirmation Trading system 9: Trend following with the TEMA Trading system 10: Bull/Bear fear Trading system 11: Simple RSI with equity curve filter Trading system 12: The range indicator (TRI) Trading system 13: Volatility

“algorithmic” from “software” has been also reflected in media studies. [2] “Drawing on contemporary media art practices and on studies of visual and algorithmic cultures, I would like to develop the idea of algorithmic superstructuring as a reading of aest

American Revolution, students are exposed to academic, domain-specific vocabulary and the names and brief descriptions of key events. Lesson 2 is a simulation in which the “Royal Tax Commissioners” stamp all papers written by students and force them to pay a “tax” or imprisonment.