R.51 Algorithmic Trading

Question 1

Execution vs HFT algorithms

1. Execution Algorithms
2. High-Frequency Trading (HFT) Algorithms

Answer 1

Execution Algorithms: break down large orders and execute over a period of time. This helps reduce probability trading strategy being detected by others and minimizes price impact. Examples include:​​

• Volume-weighted average price (VWAP): uses the historical trading volume distribution for a particular security over the course of a day and divides the order into slices
• Implementation shortfall - dynamically adjusts schedule of trade in response to market to minimize spread
• Market participation slices order into segments intended to participate on a pro-rata basis with volume throughout the course of execution period

High-Frequency Trading (HFT) Algorithms: rules for trading on real-time market data that a computer uses to pursue profit opportunities. “High frequency” refers to rapidly-updated information sources that these algorithms rely on (market data feeds and news feeds.)

Statistical arbitrage trades securities that historically have moved together. Examples include:

• *• Pairs trading** – looking for breaks in correlated relationships
• *• Index arbitrage** – looking for breaks between instrument and index of its sector
• *• Basket trading** – applying techniques to custom basket of instruments rather than individual instruments
• *• Spread trading** – usually created by taking a long and short position, hoping to profit on the spread change. This is popular in the futures market.

Intra-market spread: time or calendar spread w/ same commodity (purchase of July corn and the sale of December corn)

inter-market spread: same delivery month on two commodities that have a fundamental relationship (purchase of February lean hogs and the sale of February live cattle)

inter-exchange spread: different exchanges (purchase of March Kansas City wheat and the sale of March Chicago wheat). More complex inter-exchange multi-legged spreads include:

crack spreads (trading the differential between the price of crude oil and petroleum products),

spark spreads (trading the theoretical gross margin of a gas-fired power plant derived from selling a unit of electricity against the price of the fuel required to produce this unit of electricity, including all other costs of operation, maintenance, and capital and other financial costs), and

crush spreads (the purchase of soybean futures and the sale of soybean oil and soybean meal futures).

• *• Mean reversion** – based on the assumption that prices will revert to a long-term mean.
• *• Delta neutral strategies** – profits are made independent of the price change in the underlying asset. Stocks and options are combined to create a net delta of zero.
• *Liquidity aggregation and smart order routing** - Many venues trade the same instrument, which is known as market fragmentation. Algorithms can be used to search for the market with the best price.
• *Real-time pricing of instruments** - Techniques can be used to price securities like bonds much faster than with traditional techniques.
• *Trading on news** - Algorithms can digest news much faster than traders, thus giving an edge in market trading.
• *Genetic tuning** - Algorithms can be tested with real data before going live with trading. This can be used to determine the ones that are most likely to be profitable.

Question 2

Evolution of Algorithmics and HFT (key drivers)

Answer 2

Market fragmentation

• Many exchanges and alternative trading systems are now used, and any one system may represent only a small portion of aggregate liquidity for an instrument. Algorithmic techniques strive to maximize the aggregate liquidity. This can change quickly, so low latency is important.
• Algorithmic techniques to capitalize/address
  
  • ​Liquidity aggregation: creates a “super book” that combines liquidity on a per symbol or currency pair basis. This offers a global-ordered view of market depth for each instrument regardless of which trading venue offers the liquidity. For example, the best bid for a Eurodollar future may be on the Chicago Mercantile Exchange (CME) and the second best may be ELX Markets, a fully electronic futures exchange.
  • Smart order routing: sends the order to the relevant market(s) on which the quote is displayed. Low latency and rapid update are clearly important to avoid trading on stale liquidity information.

Other Key drivers over past 15 years include:

• Opportunities in new asset classes: Initially, equities and futures were the only assets with electronic, open, and fragmented trading. Now many other assets such as bonds offer the same opportunities.
• Opportunities in cross-asset class trading: Assets with correlation relationships can be exploited when the pricing diverges.
• Opportunities in new geographies: Algorithmic trading and HFT started in the US and UK markets, but now has spread to many more.
• Opportunities in cross-border trading: Statistical arbitrage opportunities are available when assets are listed in multiple countries.

Question 3

Risk Management uses of trading algorithms

Answer 3

Two approaches used to reduce trading risk are:

1. Real-time pre-trade risk firewall

Exposures can be calculated in real-time as trades take place. Trades that would increase the risk exposure over specified limits could be blocked. This is also useful for brokers offering sponsored access, which allows clients to trade directly using the broker’s exchange memberships.

2. Back testing and market simulation

It is beneficial to test algorithms with real historical and pre-planned scenarios before using them for live trading.

Question 4

e. describe the use of technology in risk management and regulatory oversight

Key Technologies for algorithmic trading

Risk Management Uses of Trading Algorithms

Regulatory Oversight: Real-Time Market Monitoring and Surveillance

Answer 4

Key technologies integral to algorithmic trading include:

• Execution management systems—front-end trading systems that allow access to broker algorithms as well as access to custom algorithms integrated with the EMS.
• Complex event processing (CEP)—a platform specifically designed for complex analysis and response to high-frequency data. CEP platforms incorporate graphical modeling tools that rapidly capture and customize strategies and a trading engine connected to any combination of cross-asset market data and trading venues. Used widely for algorithmic trading, HFT, liquidity aggregation, smart order routing, pre-trade risk analysis, and market surveillance.
• Tick database—a real-time time-series database designed to capture and store high-frequency market data for analysis and back testing.

Risk Management Uses of Trading Algorithms

• Real-time pre-trade risk firewall: if trade breaches a pre-established risk threshold, trades can be blocked. Can monitor for erroneous trades, such as fat finger trades (trading errors like buying 1 share at $1,000 instead of 1,000 shares at $1), and block them. This capability is useful not only for trading groups but also for brokers offering sponsored access. Sponsored access enables direct market access (DMA) for clients of the broker who want to trade using the broker’s exchange memberships.
• Back testing and market simulation: test algorithms with a variety of real historical and pre-planned scenarios before putting them to work live to see how they can be expected to perform. This process can be done in conjunction with realistic and tunable market simulators.

Regulators can better detect patterns such as:

• Insider trading - Potential insider trading can be spotted with spikes in volume by a trader in a security that is not common for that trader. Often it is followed by news events that materially impact the price.
• Front running orders - Front running occurs when a trader buys a security prior to a large buy-side order being placed. Algorithms can be used to detect these coincidental orders.
• Painting the tape - This is a practice of traders manipulating the market to move the price. For example, a trader could continuously take the best offer in the market to drive the price up prior to selling a large quantity.
• Fictitious orders - Traders can enter large quantities of fictitious orders and then cancel them. This is done to confuse other algorithms. Layering involves placing a real order on one side of the market and multiple illegitimate orders on the other side. Spoofing is done by placing limit orders that are not intended to be executed.
• Wash trading - Traders can buy and sell the same security to artificially increase the volume to drive up the demand.
• Trader collusion - Traders can work together to manipulate the market, such as the Libor rates.

Question 5

Positive and Negative impacts of algorithmic and HFT on securities markets

Answer 5

Positive impacts include:

Minimized market impact of large trades – this is the key goal of algorithmic trading

Lower cost of execution – algorithms are cheaper than traders

Improved efficiency in certain markets – arbitrage opportunities are quickly exploited and eliminated

More open and competitive trading markets – smaller teams have access to trading operations

Improved and more efficient trading venues – competitions for liquidity in trading puts downward pressure on exchange costs

Negative impacts include:

Fear of an unfair advantage - Many believe algorithmic traders have an unfair advantage of seeing market movements before ordinary investors. Also, many contend behavior such as spoofing is widespread.

Acceleration and accentuation of market movements - Since the trading involves no emotions, it can accentuate crashes.

Gaming the market

It is very easy to trick the market with automatic trades.

Increased risk profile - Algorithmic trading can increase the efficiency of traders, but it also increases the risk.

Algorithms gone wild - If the logic is less than perfect, automated trading can lead to very big losses. One such algorithm lost the trading firm $440 million in just 30 minutes. Real-time pre-trade firewalls can reduce this risk.

Potential for market denial-of-service-style attacks - Rapid orders from algorithmic trading can overwhelm exchanges.

Additional load on trading venues - Many algorithms adjust their bids and offers continuously. This puts a strain on exchanges.

Increased difficulty of policing the market - The rapid trading in various markets makes the task difficult for regulators. Dark pools are trading venues that do not publish their liquidity and are only available to selected clients.

Question 6

a. Define algorithmic trading
* what are the two types

Answer 6

Algorithmic trading

• algorithm is “a sequence of steps to achieve a goal,” and algorithmic trading is “using a computer to automate a trading strategy.”
• replicates the decisions a human trader would make
  and the orders they would place, but at speeds thousands of times faster.

Two types of trading algorithms:

1. Execution algorithms
2. High-frequency trading (HFT) algorithms.