Equity Research (Part 4) Algorithmic Trading

In today’s highly competitive and cost-conscious trading environment, fund managers and buy-side traders increasingly rely on computerized algorithms provided by brokers. Algorithmic trading often referred to as black box trading, computer trading, automated trading, program trading, basket trading, enhanced execution, or execution strategies, involves rules-based trading where a quantitative model automatically generates the timing and size of buy or sell orders based on predefined goals, parameters, and constraints. Its primary aim is to determine the optimal time for an order to be placed to cause the least amount of impact on the price of the financial instrument.

Evolution and Key Players

The evolution of algorithmic trading can be traced back to the early 1970s with the computerization of trade execution systems, notably the introduction of NYSE’s DOT (Designated Order Turnaround), which routed trading orders to appropriate venues. Program trading became widely used in the 1980s for trading in equity and futures markets, particularly in stock index arbitrage. While program trading was linked to the 1987 stock market crash, the 1990s saw the development of fully electronic trading systems. These systems initially impacted market makers’ advantages and decreased market liquidity, leading to the development of algorithms that split large orders into smaller quantities to achieve executions at lower average prices. As electronic markets advanced, strategies like arbitrage and statistical arbitrage became more easily implemented by computers, processing data faster and executing transactions more efficiently.

Some of the largest players in algorithmic trading include UBS, Citigroup, Goldman Sachs, Morgan Stanley, Credit Suisse and Deutsche Bank. Lehman Brothers, for example, introduced algorithmic trading for shares on India’s National Stock Exchange (NSE) through its LMX Trading Strategies suite.

Components of Algorithmic Trading

An effective algorithmic trading system requires several core components:

• Real time and historical market data.
• Algorithms designed to perform correlation analysis, identify trading opportunities, determine optimal timing for order launch, and measure trade execution against benchmarks like VWAP (Volume-Weighted Average Price) and TWAP (Time-Weighted Average Price).
• Order management/order processing capabilities.
• Connectivity to various liquidity pools, including exchanges, ECNs (Electronic Communication Networks) and inter-dealer brokers.
• Integration with internal systems for trading, order management, risk management, compliance, and back-office functions.

Algorithmic Trading Models

Algorithmic trading models use complex mathematical calculations to decide the timing, price and quantity of orders for execution, with the choice depending on the trader’s perspectives and objectives. Broadly these models are categorized into the following types:

1. Price Algorithms: These algorithms utilize predefined price benchmarks, such as closing price, market price, arrival price or calculated variables like VWAP and TWAP to execute transactions. VWAP and TWAP are gaining popularity due to their high execution performance.
2. Time Algorithms: These models operate primarily based on time executing a certain amount of an order during each time period or executing transactions based on time-sensitive information.
3. Implementation Shortfall Algorithms: Employing quantitative techniques, these algorithms aim to balance liquidity impact and opportunity costs. They calculate the risks and returns associated with order timing and price to ensure optimal liquidity and minimal losses from opportunity costs.
4. Volume Participation Algorithms: These algorithms execute transactions based on a predetermined limit of execution volume, ensuring it does not exceed a specified percentage of the total market volume.
5. Smart Order Routing Algorithms: These algorithms operate across multiple venues, scanning them for the best possible price to execute transactions. They are particularly efficient for large orders or transactions and are analogous to black-box models with optimized parameters for best results. Custom models also allow traders to define their own parameters.

Algorithmic Trading Process and Strategies

Algorithms are used throughout the trade cycle, typically classified into pre trade analytics, the execution stage, and post trade analytics.

• Pre trade analytics involves analyzing historical and current price/volume data to help clients determine where and when to send orders, and whether to use algorithms or manual trading. This analysis helps buy-side traders understand and minimize market impact by choosing appropriate aggressiveness levels and time horizons for trading.
• Execution Stage allows traders to create stock lists, choose strategies (e.g implementation shortfall), and set start/end times. Traders can monitor algorithm performance in real-time and adjust parameters if needed. Some algorithms in this stage sweep crossing networks and “dark books” (liquidity pools that match orders without publishing quotes).
• Post trade analytics tracks commissions and uncovers all costs from trade initiation to execution. Its purpose is to improve execution quality and inform investment decisions. VWAP is a prevalent benchmark for post-trade analysis though it is less useful for strategies not aiming to follow the market midpoint.

Important Factors for Algorithmic Trading Suitability:

• Market Cap: Most suitable for large-cap stocks generally not used for small caps.
• Average Trade Volume: Requires high trade volumes.
• Average Trade Frequency: Preferable for high trade frequency; not ideal if high volume but few transactions.
• Volatility: High volatility makes algorithmic trading less efficient.
• Spread: A greater spread indicates lower liquidity, making algorithmic trading less efficient. These factors are critical for statistical checks to determine suitable transactions for algorithmic trading.

Algorithmic Trading Strategies:

• Iceberging: Breaks very large orders into smaller ones executed over time, commonly used to reduce transaction costs. “Guerrilla” is an algorithm designed for this.
• Arbitrage: A risk-free profit-earning strategy implemented by complex statistical models often involving multiple securities or different trading venues.
• Market Making: Profits from the bid-ask spread by placing limit orders to buy below and sell above the current market price.
• Benchmarking: Algorithms are designed to follow an index’s movement and execute transactions for maximum profits. “Sniffer” is an algorithm for this also useful for identifying volatile markets.
• Gaming: Relies on understanding and exploiting the programming skills of other algo traders. “Gamers” or “sharks” use small market orders (“pinging”) to detect large, hidden (“iceberged”) orders in dark pools.
• Black Box Trading: Utilizes complex models like neural networks and genetic programming to identify patterns and trends for timing transaction execution.

Approach, Advantages, Disadvantages and Trends

Achieving best execution through algorithmic tools depends on several critical ingredients:

• A clear understanding of portfolio management strategies’ objectives.
• Robust pre-trade models.
• Balancing timing and impact cost issues.
• Effective, intelligent integration of Order Management Systems (OMS) and direct market access trading platforms.
• Close iterative relationships with algorithmic trading providers.
• Thorough post-trade analysis and feedback.

To meet these requirements, an algorithmic platform should have the following attributes:

• Adaptable: Providing high-speed transmission of market data and transaction messages offering a vendor-agnostic platform for data acceptance and distribution, and including pre-integrated security and monitoring for compliance and cost-effective operations.
• Streamlined: Ensuring optimized acquisition, processing and delivery of market data through an efficient and integrated platform.
• Reliable: Enabling continuous delivery of market data with robustness to support front office needs.
• Open Architecture: Promoting interoperability through open published specifications for Application Program Interfaces (APIs), protocols and data/file formats, allowing for flexible and reconfigurable solutions.

Advantages of Algorithmic Trading:

• Reduction in trading costs: Significantly impacts market makers profitability and decimalization of spreads, reducing traditional implicit and explicit transaction costs like floor brokerages and commissions, sometimes to zero.
• Internationalization/crossing of order flow: Smart routers identify appropriate venues for order execution, facilitating international order flow.
• Reduction in cost of entering markets: The radical fall in prices of hardware, software and networking devices has lowered technology costs, alongside increasing availability of custom made application solutions.
• Cross Asset Trading Opportunities: Traders can easily identify cross asset trading opportunities, which helps in hedging risks, e.g. identifying a derivative hedge for an equity purchase. Modern platforms provide tools to test new strategies before deployment.
• Connecting Dark Pools Creates More Liquidity: Broker-dealers use algorithms to match buy and sell orders in dark pools (off-exchange, anonymous markets) without publishing quotes, which enhances liquidity, pricing for clients, and potentially higher commissions for brokers.

Disadvantages of Algorithmic Trading:

• Lack of Visibility: Despite understanding how algorithms work, there’s a lack of visibility and transparency during order execution.
• Algorithms Acting on Other Algorithms: Algorithms can track and “reverse engineer” competitors trading patterns and trends, exploiting predictions of buy and sell orders.
• Choosing Optimal Algorithm: It’s a major challenge to select the best algorithm for a specific stock type due to a lack of tools to identify the optimal strategy and the absence of benchmarks for quality assessment.
• Timing of Using Algorithms: The same algorithm may not be optimal for a stock at all times making the timing of usage a significant concern. Buy-side firms struggle to evaluate when to use a particular algorithm, especially with liquidity and time constraints.
• More Monitoring & Testing: Requires careful real-time performance monitoring and extensive pre- and post-trade analysis to ensure proper application. Traders must track past algorithm performance to find optimal solutions for client portfolios, particularly for urgent, large-volume orders.

Latest Trends in Algorithmic Trading:

Algorithmic trading has transformed the trading landscape, giving institutional players greater control over trading styles, timings, and portfolios.

• Customized Algos: Buy-side firms are moving towards customized algorithms for greater control over transactions.
• New Trading Strategies: Buy-side firms are developing diversification strategies by trading combinations of equities, fixed income and/or derivative instruments to lower overall costs and risks.
• Additional Services by Sell-side Firms: Sell-side firms are offering services like portfolio diversification, research, Customer Relationship Management (CRM) and increased privacy/anonymity. New algorithms are being developed for various purposes including VWAP, TWAP, small-cap illiquid, mid-cap liquid, passive and aggressive strategies.
• Migration to Other Markets: Algorithmic trading is expanding into foreign exchange (Forex) and fixed income markets. It’s already adopted in major Forex markets due to their complex nature and abundant arbitrage opportunities.
• Algorithms for News Analysis: There’s demand for algorithms that can analyze news and its impact on equity markets, alerting traders to significant stock value changes based on company announcements.
• Algorithms for Risk Management and Regulatory Requirements: Due to high trading risks, algorithms are in demand to mitigate risk by monitoring various conditions (e.g. using Value-at-Risk or VaR) and to address regulatory compliance issues.
• Next Generation Algorithms: Driven by client demand and innovation, future algorithms will be able to speak to traders, providing dynamic updates and allowing interaction. Adaptive algorithms are emerging that adjust their execution in real time based on market conditions, similar to human traders.