Slide 1

Slide 1 text

Efficient risk control — challenges and techniques. ARQA's case study Vladimir Kurlyandchik director, business development

Slide 2

Slide 2 text

FLASH CRASH Introduction 1 $31.7 mln $100 mln $440 mln 2012 2013 2013 May 2010

Slide 3

Slide 3 text

Typical risk models 2 For equities:  counterparty limits  size of order  frequency of orders (in some cases) For derivatives:  fixed amount of dollars (rubles, euros) per contract including options For FX:  full prefunding for corporates  leverage 1:100 and higher for private investors For bonds:  counterparty limits

Slide 4

Slide 4 text

Client needs 3 Private investor: One account to trade any asset type HFT client: Different types of arbitrage strategies Institutional client: Effective risk management

Slide 5

Slide 5 text

Gap reasons 4  Trading platform fragmentation — broker uses different platforms for different asset types and exchanges  Risk control works as defense against fear  Risk management department is not a business but a support unit of sell-side company  Latency reasons — people don’t want to affect latency by complex risk checks

Slide 6

Slide 6 text

How to solve the problem? 5 Level 1 Changing the attitude to risk management Level 2 Clear understanding of client groups (trading patterns and requirements) Level 3 Easily adapted risk management solution architecture Level 4 Quality of implementation of risk management solution

Slide 7

Slide 7 text

ARQA approach 6

Slide 8

Slide 8 text

ARQA approach 7  In all asset types  In any currency  For different settlement dates  Flexible rules of calculation Position

Slide 9

Slide 9 text

ARQA approach 8  Different sources: • trading venues (LSE, LSE derivatives, Turquoise, CME, Eurex, EBS ICAP, WSE, MOEX) • market data vendors (Thomson Reuters, SunGard, S&P Capital IQ) • FX liquidity pools  As fast as possible Market Data

Slide 10

Slide 10 text

ARQA approach 9  Full prefunding  Several margin trading models (one leverage for all assets, different discounts for different assets, etc.)  Classical prop-trading limits (limit on short, long, net in a particular asset)  Netting opposite positions in correlated instruments (stock against single stock futures, currency pair against corresponding futures, local stock against GDR)  SPAN-like methodology RISQ models

Slide 11

Slide 11 text

ARQA approach 10  Flat fee  Fixed amount of money per contract  Commission scale  Minimum and maximum commission amount Pre-trade and post-trade calculating Fees

Slide 12

Slide 12 text

ARQA approach 11 Multi-currency  Multi-currency support  Different scenarios  Business advantages for different client groups  Full risk control

Slide 13

Slide 13 text

ARQA approach 12 RISQ terminal Main tasks  show real-time position from different angles  show real-time information about risk parameters for client position  what-if calculators Automate some procedures  margin calls  position termination  obligation rollover

Slide 14

Slide 14 text

Infrastructure cases 13 Example 1: Servicing retail clients Gateways Replicator End clients RISQ/EMS server 1 RISQ/EMS server 2 End clients Gateways Exchanges

Slide 15

Slide 15 text

Infrastructure cases 14 Example 2: Servicing HFT client Black box RISQ server Transactions RISQ filter module Yes/No Exchange Drop copy Gateway

Slide 16

Slide 16 text

Infrastructure cases 15 Example 2: Servicing HFT client Black box RISQ server QUIK KillSwitch Yes/No Exchange Gateway Trigger exchange KillSwitch

Slide 17

Slide 17 text

Infrastructure cases 16 Example 3: Extending existing infrastructure by advanced risk checks End clients Broker’s EMS RISQ server FIX adapter FIX order router Exchange Gateway Transactions Drop copy

Slide 18

Slide 18 text

Numbers 17 Capacity  Number of positions – 2 million  Number of orders – 30 million  Number of trades – 2 million Latency  RISQ server, margining on stocks with discounts – 1 ms per transaction  RISQ server portfolio margining – 2-3 ms per transaction  FIXPreTrade RISQ filter for Windows, any model – 100 mcs  FIXPreTrade RISQ filter for Linux (tcp off load), any model – 30 mcs

Slide 19

Slide 19 text

Development and testing 18  Dedicated business analysis staff  Enrichment of internal expertize by internal and external seminars  Cooperation of inside development team and technical support  Delivery new versions for testing with special cases from developers and analysts  Specially developed environment to run automated tests  Investigations of new releases before go live dates

Slide 20

Slide 20 text

No content