Automated trading

from Wikipedia, the free encyclopedia

Automated or algorithmic trading (also algorithmic trading , algo trading , black box , high frequency trading , flash trading or gray box trading ) generally refers to the automatic trading of securities using computer programs.

According to the Securities Trading Act ( Section 80 (2) WpHG) algorithmic trading is described as trading in financial instruments in which a computer algorithm automatically decides on the execution and the parameters of the order. This does not apply to systems that only confirm orders or forward them to other trading venues.

To date, no clear definition has been established in the literature of business informatics and economics . Many authors understand this to mean computer programs that are used to electronically transfer existing buy and sell orders to the stock exchange. The other group of authors understand it to be computer programs that make buying and selling decisions independently. In this context, one can distinguish algorithmic trading between buy-side and sell-side financial institutions.

history

On the development of automated trading: Exchanges report a share of up to 50 percent of sales. From 2004 to 2006, automated trading at Eurex quadrupled. In contrast, traditional trade grew only slightly. The EUREX is believed that currently about 20-30% of all sales generated by automated trading. A growth rate of around 20% per year is expected within EUREX. According to a study by the AITE Group, around a third of all securities trading in 2006 was controlled by automatic computer programs and algorithms. AITE estimates that this percentage could reach around 50% by 2010. As Gomolka points out, these figures for stock exchange turnover are to be viewed critically. This is because the exchanges only see those orders that are transmitted to the exchange by machines and that are recorded in the electronic order books (see transaction support). What proportion of the exchange turnover is generated by machines (see decision support) and what proportion is entered into the order systems by human traders cannot be measured by the exchanges.

In early July 2009, a former employee of the American financial services company Goldman Sachs was arrested by the FBI for allegedly stealing parts of the software used by the company for automated trading. According to the public prosecutor's office, the software is also suitable “to manipulate markets in an unfair way”. However, he has since been acquitted for not stealing a physical item under US law. For the most part, the programs he took with him were open source programs that he himself improved.

Algorithmic trading for placing orders

Depending on the degree of automation, the computer can independently decide on certain aspects of the order (timing, price, volume or time of the order placement). In the so-called "Sell Side Algo-Trading" (e.g. brokerages), large orders are divided into several smaller trades. This enables market impact , opportunity costs and risks to be controlled. The algorithm determines the split and the time (timing) of the orders on the basis of predefined parameters. These parameters typically use both historical and current market data. Algorithmic trading is used by brokers on the one hand for proprietary trading, on the other hand it is also offered as a service to brokers' customers (due to the complexity and resource situation, institutional investors have a certain urge to access broker solutions). The advantage of automated trading is the high speed at which they can place trades and the higher amount of relevant information that they observe and process compared to humans. This also goes hand in hand with lower transaction costs. The prerequisite for algorithmic trading is that an order or a trading strategy has already been placed. In contrast to automatic trading or quote machines, this is about intelligently distributing an order to different markets. It is not about automatically shooting offers into the market based on parameters.

Automated trading as decision support

Automated trading is used by hedge funds , pension funds, mutual funds, banks and other institutional investors to automatically generate and / or execute orders. Here, computers independently generate buy and sell signals, which are converted into orders in the financial center before people can even intervene. Algorithmic trading can be used with any investment strategy: market making , inter-market spreading , arbitrage , trend following models or speculation. The specific application of computer models in the investment decision and implementation is different. Computers can either only be used to support the investment analysis ( quant funds ) or the orders can be generated automatically and forwarded to the financial centers ( autopilot ). The difficulty with algorithmic trading lies in the aggregation and analysis of historical market data as well as the aggregation of real-time prices in order to enable trading. Setting up and testing mathematical models is also not trivial.

Differentiation between high frequency trading and systematic trading

In the literature, algorithmic trading is often equated with high - frequency trading , in which securities are bought and sold in fractions of a second. According to a study by FINalternatives, fund managers categorize the area of ​​algorithmic trading very differently. Over 60% of those surveyed understand high-frequency trading as transactions in a period of 1 s to 10 minutes. Approx. 15% of the respondents understand this to mean transactions in a period of 1–5 days. Aldridge (2009) categorizes Algorithmic Trading exclusively as high frequency trading. Gomolka (2011), on the other hand, summarizes both high-frequency trading (in fractions of a second) and systematic trading (long-term over several days) under algorithmic trading. He emphasizes that computer programs are not only used in the short term (e.g. for flash trading), but can also act independently in the long term over the course of several minutes, hours or days.

Effects on financial market stability

In contrast to the computer exchange , in which computers only serve as a communication platform for linking suitable buy and sell offers , the system automatically places such offers and looks for trading partners. You are made jointly responsible for the stock market crash on October 19, 1987 , Black Monday . Their “if-then” algorithms are said to have ensured that more and more blocks of shares were sold after prices began to fall, which ultimately led to panicked sales. On May 6, 2010, the Dow Jones fell by over 1,000 points in eight minutes. This crash was not originally triggered by a high frequency program, but a sell order from the trading house Waddell & Reed, which placed 75,000 S & P500 E-Mini futures contracts on the market within 20 minutes via market order. This flash crash prompted the SEC to tighten its circuit breaker rules, according to which future price drops of over 10% for a share should lead to an automatic suspension of trading.

Regulatory importance of algorithmic trading

The regulatory classification of a trading strategy as "algorithmic" entails numerous follow-up obligations. In accordance with Section 80 (2) of the German Securities Trading Act, algorithmic traders must fulfill labeling obligations both towards the respective trading venue and towards the Federal Financial Supervisory Authority. In addition, algorithmic traders are subject to a license requirement for banking transactions in accordance with Section 1 Para. Section 32 KWG. The legislator justifies the obligations with the dangers of algorithmic trading. However, the strict supervision of algorithmic traders has met with criticism in many places. On the one hand, the far too broad wording of Article 4 (1) No. 39 of Directive 2014/65 / EU (MiFID II) is criticized. On the other hand, it is criticized that the legislators have made the supervisory framework partially confusing and left unanswered (economically) essential questions about the distinction between algorithmic trading and manual trading. For example, the question of the number of automatically determined order parameters required to justify algorithmic trading or the question of the point in time of the existence or the design of human intervention to avert algorithmic trading have not yet been clarified.

Web links

Individual evidence

  1. ^ SEC Expected to Limit 'Flash' Trading
  2. ↑ Gomber, P., Gsell, M. & Wranik, A. (2005): Algorithmic Trading - Machines on Financial Markets in Die Bank special edition for EBIF (2005), pp. 40–45
  3. ^ A b Johannes Gomolka: Algorithmic Trading . Universitätsverlag Potsdam, 2011, p. 4–19 ( PDF 4.9 MB [accessed May 1, 2018]).
  4. ↑ The Ultimate Money Machine ( Memento of May 11, 2007 in the Internet Archive ), Iran Daily May 7, 2007
  5. ^ Spiegel.de: Goldman Sachs: Software theft reveals vulnerability of Wall Street , accessed on May 6, 2011
  6. ↑ spiegel.de: US court ruling: Program code cannot be stolen , accessed on July 1, 2012
  7. ↑ Moving markets Shifts in trading patterns are making technology ever more important, The Economist, Feb 2, 2006
  8. ↑ Gomber, P., Groth, S., Gsell, M. (2009): Trading and Ordering Behavior of Computers versus Human Traders , Journal for the Entire Credit System , Issue Technology 02 of April 15, 2009
  9. ↑ Survey by FINalternatives in July 2009 about the opinions of fund managers to categorize to time of high frequency trading (Source: Aldridge, Irene (2009), p. 22 cited after FINalternatives 2009)
  10. ↑ Aldridge, Irene (2009): High Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Strategies, cited after FINalternatives 2009.
  11. ↑ Gomolka, J. (2011): Algorithmic Trading , Universitätsverlag Potsdam, p. 175. (PDF; 4.9 MB)
  12. ↑ SEC draws conclusions from the dramatic Dow crash on May 19, 2010, Handelsblatt
  13. ^ German Federal Council, printed matter 607/12, p. 10
  14. ↑ FIA, Reply on ESMA Consultation Paper on MiFID II / MiFIR, May 22, 2014, p. 35.
  15. ↑ Dr. Volker Baas and Mert Kilic, Regulation for algorithmic exchange trading leaves many questions unanswered , in: Börsen-Zeitung, issue 71 of April 10, 2020, p. 9