Algorithmic Trading Trends and Drivers

January, 2005
______________________________________________________
Presented by:

Tellefsen Consulting Group, Inc.

AGENDA
Background Algorithmic Trading Components Trends Factors Impacting Algorithmic Trading Design Flaws of Central Limit Order Books System Design Limitations Conclusions

CG

Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

BACKGROUND
Liquidity pool hubs and spokes: Equity/Options and Futures Exchanges Equity ECNs Fixed Income Intra-Dealer Brokers Market Makers, Specialists, Proprietary Trading Firms, Trading Arcades, Scalpers/Day Traders Buy Side/Quant Trading Shops Boutiques The evolution of trading technology solutions:
ECN aggregators, smart routers, algorithmic slicers, etc.
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

BACKGROUND (Contd)

Program trading, automatic order generators and automated order routing systems are used mainly by sell-side firms to conduct business as: Principal/proprietary trading Agency/customer facilitation In: Equities, Fixed Income, Futures and Options Strategies include index substitutions, risk modifications, liquidation of exchange for physicals (EFPs), etc. Buy programs occur when the futures market is overvalued, relative to the cash market. Sell programs occur when the cash market is overvalued, relative to futures.
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

BACKGROUND (Contd)

Largest players today include: Citadel Investment Group Citigroup Credit Suisse First Boston Deutsche Bank Goldman Sachs Lehman Bros. Morgan Stanley Susquehanna Investment Group UBS
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

BACKGROUND (Contd)

Computers do at warp speed what traders used to do by hand:

e.g., Auto-quoting options across 100 price points

A large percentage of order flow arrives via computer to computer APIs. Proprietary trading systems can bombard an electronic trading system with high burst rates of orders/second. Volume created by program trading engines has become the dominant mode of trading (*now >50% average daily NYSE volume).
* - Source New York Stock Exchange.

CG

Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

ALGORITHMIC TRADING COMPONENTS

Real time and historical market data. Algorithms to: Perform correlation analysis Identify trading opportunities Determine optimal timing to launch Measure trade execution against benchmarks (VWAP TWAP, etc.) Order management/order processing. Connectivity to liquidity pools: Exchanges, ECNs, inter-dealer brokers, etc. Integration with internal systems: Trading Order Management Risk Management Compliance Back Office
Algorithmic Trading Trends and Drivers 7

CG

Tellefsen Consulting Group, Inc.

Consultants to the Financial Services Industry

TRENDS
Intense pressures on trading costs: Impact of decimalization on spreads, market maker and specialists profitability. Impact of penny jumping on institutional order flow, block trades Implicit and explicit costs - slippage/market impact, floor brokerage and commissions Reduction in soft dollar relationships. Buy-side firms focusing on trading strategies that involve a combination of equities, fixed income and/or derivative instruments that lower overall trading costs and risks when traded together.
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

TRENDS (CONTD)

More sell-side firms offering program trading functions to compete/attract more order flow. Sell side firms moving deeper into the aggregation business: Provide buy side clients with the ability to diversify risk, get comprehensive pricing, acquire research, increase anonymity, maintain relationships, etc. Integrate algo trading strategies into OMS and front end trading systems: Buy side users want to use algorithmic strategies to split up large orders into smaller ones, to reduce market impact. Algos for VWAP, TWAP, small-cap illiquid, mid-cap liquid, passive, aggressive, etc.
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

Consultants to the Financial Services Industry

TRENDS (CONTD)

Buy side firms getting access to algorithms from multiple brokers, OMS vendors. Internalization/crossing of order flow. Cost of market entry continues to get lower: Technology costs less HW/SW/NW Buy versus build, time to market issues Application solutions availability You dont need to be Goldman or Salomon any more! Market structure evolution SEC Reg. NMS reforms: NYSE Hybrid Market initiative NASDAQ/the remaining ECNs Regional equity exchanges
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

10

Consultants to the Financial Services Industry

FACTORS IMPACTING ALGORITHMIC TRADING

The velocity and volume of trading: Growth of order flow and program trading volumes The talk to trade ratio Recent Citigroup Euro MTS trade rocked the bond market Technology choke points: Systems processing/capacity planning challenges Latency of matching engines The 80-20 Rule Evolution of smart routing: ISV front end trading and order routing systems
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

11

Consultants to the Financial Services Industry

DESIGN FLAWS OF CENTRAL LIMIT ORDER BOOKS

CLOBs were designed almost 30 years ago and may not be up to the task of supporting and processing the business of the 21st century. Heavy loads occur when firms are rapidly deleting and adding orders: When a trade is executed or the price of an underlying security changes, all of the orders are cancelled and replaced, stressing the system Most quotes follow the primary market, resulting in a burst of new quotes immediately after the primary quote changes During periods of peak volume, the systems experience degradation in order turnaround.
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

12

Consultants to the Financial Services Industry

DESIGN FLAWS OF CENTRALIZED LIMIT ORDER BOOKS (Contd)

There has been a growth in the ratio of orders to fills. Additionally, order book transactions per second already exceed their original design parameters and continue to grow at a geometric rate.

CG

Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

13

Consultants to the Financial Services Industry

DESIGN FLAWS OF CENTRALIZED LIMIT ORDER BOOKS (Contd)

The first sign of system stress is the degradation in system response: Usually sub-second, the latency can begin to increase to seconds During extreme market conditions, this can degrade to minutes, as transaction volumes increase Exchanges, ECNs and ATS have met the growth in TPS by using faster processors and adding additional CPU/memory: Some exchanges and ETS have denied service to high volume API users during peak periods Some have large server farms to facilitate segmenting processing resources into increasingly fine pieces
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

14

Consultants to the Financial Services Industry

SYSTEM DESIGN CHALLENGES

The most granular CLOB design dedicates a single system thread/process to a single, individual security or series: This constraint arises from the need to guarantee a FIFO matching process Having separate threads and processes for individual securities or series can facilitate higher scaling, and for the most part does: Even with multiple securities bunched together, such as a combination of active/liquid issues and less active issues However, just as there are peaks in trading (the opening, closing, special events, etc.), there are concentrations in instruments that get traded.

CG

Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

15

Consultants to the Financial Services Industry

SYSTEM DESIGN CHALLENGES (Contd)

Algo trading strategies often have order cancellation rates of ~90%, consuming ~50% of ETS system resources. A recent study at a major U.S. Futures exchange found that 60% of all trading for that exchange was done in the nearmonth of a single financial futures contract. A similar measurement of the market opening at a U.S. regional Equities exchange pointed out that 70% of all trades (priced and executed off the primary market opening) were in the three most active stocks: The most active issue accounted for 35% of the volume Therefore, system processing power and software efficiencies need to focus on the busiest instrument: Scaling is probably only achieved with 3-5 separate processes for the actual matching engine
CG
Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

16

Consultants to the Financial Services Industry

SYSTEM DESIGN CHALLENGES (Contd)

Further issues arise by attempting to separate shared memory regions: If options strategies, spreads or complex orders are being processed, it is a technical challenge to have instruments that interact resident in separate memory pools

CG

Tellefsen Consulting Group, Inc.

Algorithmic Trading Trends and Drivers

17

Consultants to the Financial Services Industry

CONCLUSIONS
The next generation of matching engines needs a rethinking: Design of the central limit order book Support for higher volumes of trades and quotes Exchange rule impact/changes thereto Market structure changes must go hand in hand with technology changes: Quote structure changes in the face of raising autoquote volumes because of narrow spreads Filtered quote streams for pro traders Mechanisms for price improvements without penny jumping
Tellefsen Consulting Group, Inc.

CG

Algorithmic Trading Trends and Drivers

18

Consultants to the Financial Services Industry