Transcript Algorithmic Trading at GMO

ProActive for Algorithmic
Trading
NEHAL PATEL
GMO
OCTOBER 21, 2008
Algorithmic Trading at GMO
 What we do: Introduction to our business
 High-level grid/distributed computing goals
 ProActive
What we do
 Our Business: Invest money for clients
 Methodology: “Algorithmic Trading”
 Computers
 Math
 (some ) Finance
 Our Team: 12 members based in Boston and Minsk
 Educational background: Math, Physics, Computer Science
 Work Experience: Finance, Academics, Biotech, Internet
startups
Current Environment
 Trades Currencies and futures on 24 hour,
continuous basis
 ~4000 trades a day, $10 billion, 9 banks
 Two data centers, ~200 nodes, dozens processes
Motivation
 Basic Problem: How can you make money using
computers for investing?
 Basic Wisdom


Trade frequently, but smartly /Law of Large Numbers
Ideas last for months not years


Agility, Creativity
Find the right edge
Math and CS are equally important
 Embrace complexity, don’t sacrifice robustness

Methodology
 Mathematics
 Machine Learning/Data mining (Like Biotech, Web)
 Dynamic Programming (Like Airline Scheduling)
 Computer Science
 Java (~750,000 lines)
 Emphasis on infrastructure
 Utilize open source
 Data Sets: 4-5 Terabytes
 Real-time, Event processing: ~30K messages/s
Algorithmic Trading at GMO
 What we do: Introduction to our business
 High-level computing goals
 ProActive
Distributed Computing Tasks
 Real-time Stochastic Control
 Complex Event Processing
 Real Time Statistical Analysis
 Data Mining Large Datasets
 Batch Processing
 Historical Simulation
Distributed Computing Goals
 Style
 Distributed Objects living in a sea of multicast data
 Lightweight
 Agile
 Features
 Scalability
 P2P / Data Push-Pull
 Same Code both simulation and production
Algorithmic Trading at GMO
 What we do: Introduction to our business
 High-level distributed computing goals
 ProActive
Attractive Features
 POJOS
 Separation of Concerns
 High Speed simulation
 Toolset Focus
 Rapid Deployment
 Dynamic byte code
ProActive
 Consulting with ActiveEon
 Low Level Architecture
 Topics:
 Performance
 More Transparency for POJOs