The War of the Algorithms Emerging Trends in Algorithmic Trading Dr John Bates Co-Founder & Chief Technology Officer Apama VP Event Processing Products.

The War of the AlgorithmsEmerging Trends in Algorithmic

Trading

Dr John BatesCo-Founder &

Chief Technology OfficerApama

VP Event Processing Products

Progress Real-time Division

Agenda

• Emerging trends & requirements – Commoditization vs “Build Your Own”– Algorithmic War– Cross-Asset Class Algorithmic Trading

• Next Generation Algo Trading– Algo Trading Engines– Flexible Connectivity– Rapid Strategy Modelling Tools

• The Future– Self-evolving Algorithms

• Conclusions

The Interest in Algorithmic Trading

• Buyside– Competitive advantage

• Capitalize on opportunities before competitors

– Leveraging Traders’ skills• Scale each trader

– Cost advantages

• Sellside– Increase trading

volume– Attract & retain

customers

Some Statistics

• TowerGroup– “…continued growth in total

algorithmic trading, with volume doubling through 2006 and algorithmic trading initiated by the buy side tripling during the same period”

• ITG– “Algorithmic trading in use in 60% of

US buy-side firms and this percentage is set to grow”

– “The take up in Europe is currently thought to be about half of that in the US, but is expected to rise dramatically”

Black Box Strategies

• Most common way to algo trade

• Easily accessible through FIX and buy-side OMS

Instrument

Quantity

Num slices

Start time

End time

VWAP

Buy/Sell

Risks of Commoditization

• Black Box Strategies– If everyone has the same black boxes =

cancels out competitive advantage– Limited scope to use your skills – can only

parameterize– Often there isn’t a module that offers

exactly the algorithm required– An algorithm is tied to a particular broker– Can be expensive

• Pressure to differentiate– Hard for buy-side to understand what makes

one broker’s strategies better than another’s

– Fixed capability modules are too inflexible – pressure to offer cost-effective customization

– Buy-side want to know “how it works”

• Build your own– Takes a huge amount of time

& effort (IT cycle)– Maintenance issues

• Markets are continually evolving– First mover gets the

advantage– Lost opportunity cost of slow

evolution

Build Your Own

Algorithmic War

• Algorithms need to continually evolve– Competing with other algorithms

over current opportunities– New opportunities emerging– Avoid being reverse engineered– Opportunities may disappear

• Evolve or perish!

Cross-Asset Class Strategies

• Interest is growing for algorithmic trading in multiple asset classes– Equities, Futures, Forex, Bonds– Trading and Market-making (e.g.

bond pricing)

• Strategies should also be able to combine multiple asset classes– Example:

• Buy an equity, hedge with a future• Wave trade the equity

– Slice volumes based on historic volume profile

– Time slices into market based on a slightly random wave period

• Take foreign exchange position if equity is in different currency

Summary of Emerging Requirements

• Desire for competitive differentiation through strategy customization

• Desire to achieve this customization rapidly to capitalize on opportunities– Survive in the algorithmic war

• Desire to support algorithmic trading across multiple asset classes

A Plug-and-Play Approach

• Some proven Pre-trade Analytic strategies

– VWAP– Pairs– Index Arbitrage– Basket– Spread

• Combined with order management strategies

– Wave Trading/Iceberg

– Out of Market Limits

– Active re-pricing– Timeout if not

Filled– Smart Order

Routing

FeedbackFeedback

An institution’s strategy is likely to combine known analytic & order

management strategies + their secret ingredient

Need Algorithm Hosting Environment

• Hosting environment for strategies– Describe a strategy & upload into algorithmic

trading engine– Enables strategies to be easily & rapidly

created and/or extended• Efficient strategy execution

– Exploit latest “complex event processing” logic

– Plumbed directly into any number of market data feeds & order management systems

Trading Strategies

Data Feeds, e.g. Market Data, News

Actions, e.g. Place Order

Access All Markets

• Need extensible integration architecture to plug into any Exchange, OMS, Middleware etc.

• Abstract underlying connections, enabling strategies to be– Exchange-Independent– Asset Class Neutral– Backtested with simulators or historical data

IntegrationFrameworkIntegrationFramework

FIXFIXReutersReutersSonic MQSonic MQ GLGL

Business-Focussed Modelling• Enable business user to

compose, deploy, evolve and manage algorithmic strategies

• Business-focussed modelling of strategies – so users can “go inside” and customize

• Generate/evolve strategies in hours rather than weeks

• Upload directly to algorithm hosting environment

Modelling Trading Strategies

Business

User

Orders Filled

Orders Timed Out

Orders Placed

Monitor Spread Need to be

able to define strategy

process flow

Modelling Trading Strategies

Business

User

VWAP EMA

MACD

P&L

Basket

Spread

Price Feed

Inst1

Inst2MySpread

Need to be able to plug

in analytics & data sources

Spread

Orders Filled

Orders Timed Out

Orders Placed

Monitor Spread

Modelling Trading Strategies

Business

User

Orders Filled

Orders Timed Out

Orders Placed

Monitor Spread

Price Feed

Inst1

Inst2

Need to be able to define trading rules

Spread

WHEN

THEN

Spread is greater than Threshold

BuyLeg place order

SellLeg place order

Move to state [Order Placed]

MySpread

Event ManagerEvent Manager

Reuters AdapterReuters AdapterFIX AdapterFIX Adapter

ODBC AdapterODBC Adapter

Tier 1 Futures Broker

+ Apama Strategies + Scenario ManagerExisting portal

Database Server for Historic data access & storage and Back testing

Fix Gateway

“Advanced Trading”

Clie

nt

view

Ser

ver

sid

e

UK Hedge Fund – Forex Trading

• Routing orders across multiple FX liquidity pools based on price and availability

EBS Adapter

EBS Adapter

Hotspot AdapterHotspot Adapter

Risk Model

FXStrategy

user tools

Tier 1 Investment Bank - Bond Pricing

• Highly competitive pricing of bonds across inter-dealer and client networks

Historic ClientPerformance

Client-facing Bond Market

Inter-dealer bond markets

Inter-dealer market Adapter

Inter-dealer market Adapter

Real-time Trade Dashboards

Internal middleware

Adapter

Internal middleware

Adapter

Bond Strategy

Target Exposure

QuotingEngine

Start of day & intra-day positions

Market data

Price Adjustments

Market data

Price Adjustments

The Future

• Algorithms will not replace humans– They just help trading

groups scale their own capabilities

• Self-Evolving Algorithms– 1000s of permutations of the

same algorithm•All slightly different•All with simulated P&L

– Swap in most profitable

Summary

• To compete successfully in the “war of the algorithms”, algorithmic trading systems must support– Rapid evolution & customization of

strategies– Cross-asset class strategies– Support for business users to do this

• Advantages can be realised with architectures involving– General purpose algorithm hosting

engine– Integration to all data points– Business focussed environment to

compose, deploy and manage strategies

Questions?