Business, Legal & Accounting Glossary
Algorithmic trading involves entering trading orders through electronic trading platforms using an algorithm -which helps in taking into account order’s characteristic like timing, price, and the volume of the order. Besides, in many cases, algorithmic trading also initiates orders without human interference.
Algorithmic trading helps in dividing large trades into numerous smaller trades making it easier for traders to manage risks and the “market impact”. Hedge funds, pension funds and buy-side (focused on investing) institutional traders tend to use algorithmic trading to a large extent.
Also, sell-side traders such as large brokerages firms, and some hedge funds extensively use this trading mechanism, helping provide liquidity to the market along with maintaining a high level of efficiency as orders are generated and executed automatically.
There is one special class under algorithmic trading called as “high-frequency trading” (HFT). Unlike human traders, who are not very prompt processing the observed information, HFT allows computers to take complicated decisions based on information received electronically, in a fraction of seconds.
This technology has created a remarkable transformation in market microstructure, especially when it comes to providing liquidity to the system.
Algorithmic trading can be used in any investment strategy along with “market-making”, inter-market spreading, arbitrage, or just market speculation which includes monitoring of trends in desired investment vehicles.
Algorithmic trading enables enhancing both investments decisions and executions at any stage even as it can operate completely mechanically reacting as per the program fed inside the computer.
According to a data provided by Boston-based financial services industry research and consulting firm Aite Group, by 2006 one third of stocks that traded on EU and U.S. stocks market used automated programs or algorithms.
In 2009, 73% of the total U.S. equity trading volume was executed by HFT firms.
In the London Stock Exchange over 40% of all trades were fed on algorithms while in 2007 it was estimated at 70%. Globally, the share of algorithmic trades is generally found higher both in American markets and European markets.
Besides, equity trading, foreign exchange markets also use algorithmic trading. In 2006, 25% of the total foreign exchange trades were driven by algorithm programs.
Derivatives like futures and options markets can also easily integrate into algorithmic trading. By 2010, about 20% of the total volume of options trading was generated by the computer. In the same vein, bond markets are also switching towards algorithmic traders.
However, one of the concerns that surround HFT is that how much profitable it is? According to a report which was released in August 2009 by the TABB Group, a financial services industry research firm, 300 securities firms and hedge funds that used algorithmic trading earned approximately US$21 billion in profits in 2008.
Nonetheless, algorithmic and high-frequency trading drew lots of criticism for May 2010 Flash Crash. According to Security and Exchange Commission (SEC) and the Commodity Futures Trading Commission, algorithms contributed to the market volatility when DJIA plunged over 9% in few minutes, its 2nd worst intra-day swing ever to date, due to a technical glitch on May 6, 2010. However, prices quickly recovered.
Later in July 2011, International Organization of Securities Commission (IOSCO) reported that although HFT and algorithms enable traders to execute large trades along with managing skills, their usage was one of the contributing factors in the flash crash of May 6 2010.
Financial markets received a great boost in 1970s thanks to some landmark developments in technology such as computerization of the order flow. While New York Stock Exchange’s “designated order turnaround” system (DOT, and later SuperDOT), helped routing orders automatically to the proper trading post, the “opening automated reporting system” (OARS) assisted the trading experts in determining the market clearing opening price (SOR; Smart Order Routing).
New York Stocks Exchange defines Program Trading as a buying or selling orders concerning 15 or more stocks valued at above $15 million in total. In fact, it shows that all program trades are entered with the help of a computer. S&P500 equity and futures markets extensively used program trading between them in the 1980s.
During this period, some more technological inventions like “Portfolio Insurance” were designed that helped creating a synthetic put option on a stock portfolio. Computer model based on the Black–Scholes option pricing model made it possible to trade stock index futures at lightning speed.
Both these trading strategies often dubbed as “program trading” were severely criticized by many people including the Brady Report, for aggravating or even starting the stock market crash of 1987. However, the impact of program trading/computer trading on stock market crash from time to time is a matter of great debate even as it is widely discussed in academic circles.
Following the success of “program trading” in the U.S., financial markets with completely electronic execution together with similar electronic communication networks developed across the world in the late 1980s and 1990s.
Meanwhile in the U.S., decimalization, which altered the smallest tick size from 1/16 of a dollar (US$0.0625) to US$0.01 per share, may have encouraged algorithmic trading. Decimalization helped in changing the market microstructure by making it possible to have smaller differences between the bid and offer prices, thereby shrinking the market-makers’ trading advantage, and creating more liquidity in the market.
Consequently, amid higher level of liquidity, institutional investors started splitting up of orders according to algorithms so as to complete orders at better price average. These average price benchmarks are calculated by computers by using the “time-weighted average price” or more often though the volume-weighted average price.
The year 2001 saw further support for switching on to algorithmic trading in the financial markets. Just then, a team of IBM researchers published a research paper at the International Joint Conference on Artificial Intelligence where they proved in experimental laboratory versions of the electronic auctions used in the financial markets that how two algorithmic strategies (IBM’s own MGD, and Hewlett-Packard’s ZIP) could time and again do better than human traders.
MGD, a customized version of the “GD” algorithm, was invented by Steven Gjerstad & John Dickhaut in 1996/7, while the ZIP algorithm was invented at HP by Professor Dave Cliff in 1996.
In their term paper, the IBM team wrote that the financial impact from the research showing MGD and ZIP performing better than human will translate in billions of dollars annually.
As more and more electronic markets replaced manual trading, other new algorithm strategies were also introduced in the financial markets.
Algorithm strategies can be easily implemented in computers because these machines are made as such where its reaction time to temporary mispricing along with its ability to examine prices from different markets is very fast. For instance, Stealth invented by Deutsche bank, Sniper and Guerilla by Credit Suisse and hosts of other algorithmic strategies like arbitrage, statistical arbitrage, trend following and mean reversion.
These trading strategies are so fast and efficient that they started driving a new demand for Low Latency Proximity Hosting and Global Exchange Connectivity. Over here, it is important to know what Latency is while traders put together a strategy for electronic trading. Latency means the delay between the transmission of information from one source (or sender) and the reception of the information by the other end or receiver.
Latency has as a lower bound which matches the speed of light; approximately 3.3 milliseconds per 1,000 kilometers of optical fibre.
A chance of greater Latency increases when signal regenerating or routing equipment is introduced instead of speed-of-light baseline.
Trend Following: Trend following is a trading strategy which capitalizes on long-term, middle-term and short term market fluctuations that occur in different markets. The purpose of this strategy is to take advantage of a market trend on both sides, that is, going long (buying) or going short (selling) in the stock market. The idea is to benefit from constant highs and lows of the stocks and futures markets.
Traders who adopt trend following trading approaches can use techniques like current market price calculation, 20/50-200 days moving average and channel breakouts to easily identify which way the market is heading for the day and gather trade signals.
However, traders employing this strategy do not focus on price forecasting; they only commence trading when trend appears to have started and exit a trade once trend appears to have finished.
Also known as Pairs Trade, this type of trading allows traders to take profits from practically any market conditions. Accordingly, it is also called as market-neutral trade since it takes advantage from uptrend, downtrend, and sidewise movements. This strategy is characterized as a statistical arbitrage and convergence trading strategy.
Within finance, delta neutral strategy is described as a portfolio of correlated financial securities, where the value of the portfolio remains unaffected due to small fluctuations in the value of the underlying security.
This kind of portfolio normally includes options and their corresponding underlying securities in such a way that positive and negative delta components offset, resulting in the portfolio’s value being rather invulnerable to changes in the value of the underlying security.
In economics and finance, arbitrage is a technique of taking benefit of difference in prices of between two or more markets. Arbitrage trading involves identifying a combination of matching deals that benefits from the market imbalance, and the profit being the difference between the market prices.
Arbitrage Trading can only occur when one of the three conditions is satisfied.
However, arbitrage trading does not mean buying a commodity in one market where the prices are low and selling it in another market where the prices are high. The quintessence of arbitrage trading is that transaction must be executed simultaneously to cut the exposure to market risk, or the risk that prices might fluctuate on one market before both transactions have ended.
In general, this is only possible with securities and financial products which can be traded electronically. Again, it is imperative that each leg of the transaction is executed before the prices in the market may have altered.
Missing out on one of the legs of the trade and consequently having to trade at a worse price is called as ‘execution risk’ or more explicitly ‘leg risk’.
In arbitrage trading, no market risk or uncertainty should be involved.
For instance, consider this: Normally, a commodity sold at one market for a given price should fetch the same price elsewhere. However, the essence of arbitrary trading is that a good could not have same price at two or more markets even as profit comes from a difference in prices.
Traders selling wheat for example may find that price of wheat is lower in agricultural regions, and to take advantage from higher prices in the cities, they buy and transport it different cities. However, traders ignore the fact that transporting from one place to other involves transportation costs, labor cots, insurance costs and storage cost. This is not a ‘True arbitrary trading’. A true arbitrary trading is where there is no risk such as securities trading on more than one exchange. Over here arbitrage occurs by simultaneously buying in one market and selling on the other market.
Mean reversion is trading which employs mathematical methodology for stock investing; however. it can also be used for other processes. The basic concept of mean reversion trading is that stocks prices ups and downs are temporary phenomenon and the stocks will tend to have an average price during the course of time.
In mean reversion, the first task is to identify the trading range for a stock; subsequently the average price is calculated using analytical techniques linked to earnings, sales, and assets and so on.
The stock is regarded as attractive for purchase when the average price is higher than the current price, on the hand when the average price is lower than the current price the market and stock is expected to fall. In other words, a variation from an average price is expected to bring back the stock towards the average.
In general, the standard deviation of the most recent 20 prices are used a buy or sell indicator.
Besides, several stocks reporting services such as Google finance , Yahoo finance, MarketWatch, MorningStar, etc regularly provides moving averages of 50, 100, 200 days. Even though reporting services provide moving averages, it is imperative to closely look at the highs and lows for the study period.
This is a kind of arbitrage trading of small price gaps formed by bid ask price spread. Scalping traders try to act as a usual market makers or specialists. That is, buy at the bid price and sell at the ask price, the profits being the difference between the bid and ask price. This method allows for profit even when the bid and ask remain unchanged, as long as there are traders who are willing to accept market prices. It generally requires setting up and liquidating a position quickly, typically within minutes or even seconds.
The function of a scalper is more like a market makers or specialists, maintaining liquidity and the order flow of the market. A market maker is, in essence, an expert scalper. The volumes traded by a market maker are fairly large compared to a normal individual scalper. In order to closely observe a trading activity, sophisticated trading systems are employed by market makers.
Nevertheless, market makers are restricted by stringent exchange rules even as the individual trader is not. For instance, each market maker in NASDAQ is required to place at least one bid and one ask at a certain price level, in order to uphold a two-sided market for each stock represented.
A good number of trading strategies described as algorithmic trading (along with algorithmic liquidity seeking) come under cost-reduction category. Thanks to these trading strategies large trading orders are easily broken into smaller orders and entered into the market sooner or later. This fundamental strategy is known as “Iceberging”. To ascertain the success of this strategy, average purchase price is divided by the “volume-weighted average price” for the market related to a particular period.
One of the algorithms which help in finding the hidden orders or icebergs is called as “Stealth”. Nearly all of these trading strategies were first documented by ‘Optimal Trading Strategies’ by Robert Kissell.
Lately, high-frequency trading (HFT), which includes an extensive set of buy-side as well as market making sell-side traders, has turned out to be more prominent but with a certain amount of criticism.
These algorithms or designs are usually given names such as “Stealth”, created by the Deutsche Bank, “Iceberg”, “Dagger”, “Guerrilla”, “Sniper”, “BASOR”, all of them constructed by Quod Financial and “Sniffer”. However, the foundations of all these programs are based on simple mathematical models.
Dark pools are sort of substitute electronic stock exchanges where trading will take place anonymously, with most orders hidden or “iceberged.”
Gamers or “sharks” scan out big orders by “pinging” small market orders to buy and sell. When quite a few small orders are filled, the sharks may have detected the presence of a large “iceberged” order.
Commenting over the rapid speed at which new algorithms are made, Andrew Lo, director of the Massachusetts Institute of Technology’s Laboratory for Financial Engineering, once said, “now it’s an arms race.” Since lots of algorithm developers have entered the market building more sophisticated programs, competition has increased while the profits have reduced.
However, one of the inadvertent undesirable impacts of algorithm trading has been a dramatic increase in the volume of trade allocations and settlements, in addition to the transaction settlement costs linked with them.
Since 2004 though, in order to keep rising costs at check, there have been quite a few advancements in technology and services provided. For instance, individuals like Scott Kurland, have constructed solutions for combining trades executed across algorithms.
Although, in the U.S. high frequency trading firms represented only 2% of about 20,000 firms operating in 2009, it still accounted for 73% of the total trading volume. By the end of the first quarter in 2009, the total amount of assets under management of hedge funds employing HFT strategies stood at $141 billion-which was about 21% lower from their previous high. Renaissance technology was the first highly successful firm in introducing HFT strategies. By 2007-2008, high frequency funds started to gain immense popularity. Most HFT firms are market makers, ensuring enough liquidity in the market and lowering the volatility. Besides, HFT also helps in narrowing down “Bid Offer Spreads”, thereby making both trading and investing cheaper as well as easier for the market participants.
However, HFT has been a under the spotlight in the recent past following statements from U.S. Securities and Exchange Commission and the Commodity Futures Trading Commission that both algorithmic and HFT were contributing factors in the May 6, 2010 Flash Crash.
Some of the biggest names in HFT trading include: GETCO LLC, Jump Trading LLC, Tower Research Capital, Hudson River Trading as well as Citadel Investment Group, Goldman Sachs, DE Shaw, RenTech.
High-frequency trading is like quantitative trading, which is typified by short portfolio holding periods.
Under HFT strategy there are four key groups:
All portfolio-allocation decisions are taken by computerized quantitative models. HFT strategies are immensely successful because they are mainly driven by their ability to concurrently process volumes of information, a task which was not possible by ordinary human traders.
Typically high-frequency trading can be characterized by several distinguishing features
Market making is a collection of HFT strategies that helps placing a limit order to sell (or offer) higher than the current market price or a buy limit order (or bid) lower than the current price with the aim of benefiting from the bid-ask spread.
One of the most prominent market maker is Automated Trading Desk. Citigroup bought this market maker in 2007 and it accounts approximately 7% of total trading volume both at the New York Stock Exchange (NYSE) and NASDAQ.
There’s one more collection/set of HFT strategies, also known as classical arbitrage strategy. It includes quite a lot of securities such as covered interest rate parity in the foreign exchange market which helps determining the relation between the prices of a domestic bond, a bond denominated in a foreign currency, the spot price of the currency, and the price of a forward contract on the currency.
If the market prices are adequately different from those indicated by in the model to cover transaction cost, in that case four transactions can be made to guarantee a risk-free profit.
HFT makes it possible executing similar arbitrages using models of higher complexity involving several more than 4 securities.
According to TABB Group estimations, annual total profits of low latency arbitrage strategies at present, stands above US$21 billion.
A variety of statistical arbitrage strategies have been constructed whereby trading decisions are made on the basis of variations from statistically important relationships. Just like market-making strategies, statistical arbitrage can be useful in all asset classes.
It’s a highly complex program, having a subset of risk, merger, convertible, or distressed securities arbitrage that seeks to capitalize from a specific event, such as a contract signing, regulatory approval, judicial decision, etc., to alter the price or rate relationship of two or more financial instruments and makes it possible for the arbitrageur to make a profit.
Merger arbitrage is also known as risk arbitrage is a perfect example of this. Merger arbitrage usually includes of buying the stock of a company that is the target of a takeover while shorting the stock of the acquiring company.
In general, the market price of the target company is lower than the price bid by the acquiring company. The spread between these two prices centers on usually over the probability and the timing of the takeover being completed in addition to the existing level of interest rates.
The bet in a merger arbitrage is such that a spread will in the end will be zero, if and when the takeover is finished. The only risk over here is when the deal “breaks” and the spread widens extraordinarily.
Very often people tend to confuse the low latency trading with the high-frequency trading. While high-frequency trading includes computers that execute trades within milliseconds, or “with exceptionally low latency” in the terminology of the trade, Low-latency traders depend on ultra-low latency networks.
Traders using low latency trading benefit from receiving the information, such as competing bids and offers, to their algorithms microseconds earlier than the other players in the market who don’t use it.
The advancement in speed has led to the need for firms to have a real-time, collocated trading platform so as to benefit from implementing high-frequency strategies.
Strategies are continuously changed to easily discern the subtle changes in the market as well as to manage the risk of the strategy being reverse engineered by rivals.
There is also an immense pressure to constantly include new features or enhancements to a particular algorithm, such as client-centric alterations and a range of performance-enhancing changes that include, target trading performance (benchmarking), cost reduction for the trading firm or a range of other implementations.
Due to ever-changing nature of algorithmic trading strategies, traders should be able to adapt and trade intelligently under any kind of market conditions. It is imperative for traders to employ flexible trading strategies, enough to combat a vast combination of market scenarios. As a result, a significant proportion of net revenue from firms is spent on the R&D of these autonomous trading systems. Accordingly, traders spend a significant proportion of the income on R&D for upgrading various trading systems.
The majority of the algorithmic strategies demand constructing of modern programming languages, even though some traders still use strategies designed in spreadsheets.
More and more, the algorithms implemented by large brokerages and asset managers are designed according to the FIX Protocol’s Algorithmic Trading Definition Language (FIXatdl), allowing firms to receive orders to identify precisely how their electronic orders should be expressed.
Trading orders constructed using FIXatdl can then be sent out from traders’ systems via the FIX Protocol. Whereas basic models can work very well on linear regressions, more complex game-theoretic and pattern identification or extrapolative models are needed to build trading models.
For the construction of these models, developers use “Neural networks” and “genetic programming”.
Issues Surrounding Algorithm Development
Algorithms are used extensively as its use offer hosts of benefits such as improving the liquidity in the market and increasing the productivity; however, it has also drawn sharp criticism from human brokers and traders facing stiff challenge from the technology.
In addition, algorithms have been criticized for following reasons:
Generally, traders have insightful judgments of how the global economy and market works.
However, some experts find these strategies baffling. They argue that when these sophisticated systems are implemented which involves entering some numbers, and something comes out the other end, then it’s not always discernible why the “black box” highlighted certain data or relationships.
The Financial Services Authority has been maintaining close vigilance on the development of “black box” trading. In its annual report the regulator observed that there were no doubts over enormous benefits arising out of the new technology such as bringing efficiency and creating liquidity in the market. Nonetheless, it also cautioned that that growing reliance on sophisticated technology and modelling brings some greater risks like systems failure which can result in business interruption.
UK’s former Treasury minister Lord Myners once pointed out that companies could turn into the “playthings” of speculators as a consequence of automatic high-frequency trading.
Lord Myners warned that the process threatens ending the relationship between an investor and a company.
Other concerns consist of the technical problem of latency or the delay in receiving quotes to traders, security and the likelihood of the entire system going down, leading to a market crash.
Such has been shift in the market dynamics that investment banking giant Goldman Sachs spends tens of millions of dollars on this high-frequency trading strategies. The bank employs more people working in their technology department than people on the trading desk.
Algorithmic and HFT were blamed for contributing to market volatility during the May 6, 2010 Flash Crash, a day where the Dow Jones Industrial Average plummeted almost 600 points only to rebound those losses within minutes.
Amid rapid advancement in algorithmic trading, financial reporting firms have started formatting business news in such a way that it is possible to trade in algorithms after studying the news. Some of these financial reporting firms include, Bloomberg, Thomson Reuters, and Dow Jones.
In algorithmic trading, it is essential to assign significantly more constraints compared to a traditional market. A trader on one end or the “buy-side” must allow the trading system, also known as an “order management system” or “execution management system” to constantly recognize the multiplying flow of new algorithmic order types.
The R&D and other associated costs related to design complex new algorithmic orders types, along with the maintenance and development of new infrastructure, and marketing costs related to the distribution, are fairly large.
However, for complete trade execution, it was important that marketers, or the “sell-side” could express algorithmic orders automatically such that buy-side traders could just relay the new order types into their system and be prepared to trade them without constant coding custom new order entry screens each time.
A trade association called FIX Protocol LTD circulates free, open standards in the securities trading area.
The FIX language was initially crafted by Fidelity Investments, and the association Members include nearly all large and many midsized and smaller broker-dealers, money centre banks, institutional investors, mutual funds, etc.
These institution leads standard setting in the pre-trade and trade areas of security transactions. Earlier in 2006-2007, quite a few members formed an association and published a draft XML standard for communicating algorithmic order types.
The standard is known as FIX Algorithmic Trading Definition Language (FIXatdl). The first edition of this standard, 1.0 failed to take off due to limitations in the design, but the subsequent version, 1.1 (released in March 2010) is expected to be adopted extensively. The second edition is expected to dramatically reduce time-to-market and costs linked with distributing new algorithms.
Before the mid-nineties, practically all trading of securities was traded over the phone, but with the arrival of FIX, trading moved progressively towards the electronic means. The FIX protocol is employed to correspond between sell-side and the buy-side Order Management Systems (OMS) to swap orders and order execution information with no human involvement, by means of standardized messages and workflows that are described by the protocol.
To begin with, sell-side firms simply offered a contact to their ‘trading desks’ via FIX, which meant that once an order arrived at the sell-side broker, it was managed by a human trader, at least at the beginning of its lifecycle. Consequently, sell-side firms begun to provide straight access via FIX to the exchanges/markets they were members of; this is called as direct market access (DMA). Just then, a majority of sell-side firms ran their own proprietary systems to trade mechanically in the market, by means of algorithmic trading strategies, and during the course of the time they began to find that offering access to these trading strategies to the buy-side was a way to draw business and boost revenue.
Even as FIX is an extensible protocol, there were two blocks that stood in the middle as a result of sell-side firms providing access to their algorithmic trading strategies through FIX. The first challenge was that each sell-side strategy had its own parameters that had to be integrated as part of the order, so every firm ended up calling for a different set of fields (known in FIX as “tags”) to be incorporated in the FIX message. This made life very complex for the buy-side, and especially for their suppliers as adding new algorithms to their trading systems and running all the different combinations of tags became a considerable costs for their development operations.
The second concern for the market was that each sell-side firm had a precise way they wanted their algorithms to be showed on the buy-side OMS, with controls in the user interface arranged logically for easy order entry. Once more this created a challenge for the buy-side systems vendors, as every new screen for each sell-side broker needed committed development and testing attempt.
The technical designs of algorithms are not standard. In theory, a design can be characterized into three coherent units.
Amid wide usage of social networks such as Facebook, Twitter etc, some systems do implement scanning or screening technologies to decipher user’s posts. The purpose is to extract human sentiment which can influence the trading strategies.
Although invention like decimalization helped in decreasing trade sizes, further developments in algorithmic trading has made it possible to shrink trade sizes even further. Tasks previously done by human traders are being carried out computers. The lightning speed of computer connections, represented in milliseconds and even microseconds have become an essential part of the trading.
Highly computerized markets such as NASDAQ, Direct Edge and BATS, in the US, have grabbed the market share from less mechanized markets such as the NYSE.
Thanks to economies of scale arising from electronic trading, costs have dropped significantly as commissions and trade processing fees have narrowed. Besides, it has also encouraged international mergers and consolidation of financial exchanges.
Meanwhile, Competition is intensifying among exchanges with every exchange developing infrastructure for fastest processing times to complete trades. For instance, in June 2007, the London Stock Exchange started a new system called TradElect that guarantees an average 10 millisecond turnaround times from placing an order to final confirmation and simultaneously it can process 3,000 orders per second.
Not just trading strategies, competitive exchanges have also persistently tried to reduce latency with turnaround times of 3 milliseconds available. This development holds enormous significance for high-frequency traders because they have to try to identify the reliable and credible performance ranges of given financial instruments.
These professionals are time and again developing different versions of stock index funds like the E-mini S&Ps because they look for reliability and risk-mitigation along with top performance. They must sort out market data to work into their software programming so that there is very low latency and high amount of liquidity at the time for placing stop-losses and/or taking profits.
Amid increasing volatility in markets nowadays, lower latency can pose a great threat to trading where a small error can lead to a huge loss.
To help you cite our definitions in your bibliography, here is the proper citation layout for the three major formatting styles, with all of the relevant information filled in.
Definitions for Algorithmic Trading are sourced/syndicated and enhanced from:
This glossary post was last updated: 21st April, 2020 | 4 Views.