If you do not change browser settings, you agree to it. Learn more
The EU cookie law (e-Privacy Directive)
Please visit ICO website for more information. http://ico.org.uk/
|This email contains graphics, so if you don't see them, view it in your browser.|
Connect-World, the information and communication technology (ICT) decision makers' magazine. We are the decision makers' forum for ICT driven development.
|Connect-World's eLetter February I||25th February 2013|
The algo wars!
ICTs, electronic exchanges, high-frequency trading, quants and algorithms
The high drama of Wall Street trading, seen so many times on film and TV, isn’t quite what it used to be.
Today the big trades are mostly flashed through computers and closed before anyone can even breathe a sigh of relief. The algorithms - called algos by their creators - that control the trading just keep shuffling their 0’s and 1’s without a hint of trading floor excitement; even ‘quants’, allied with rocket scientists, master geeks and hackers can’t code triumph and despair. The high-tension trading room and exchange floor dramas featured in movies and TV are disappearing.
Super computers, short, really short, optical fiber cables and algorithms created by super programmers with seven figure salaries, not movie type traders, now rule. NYSE/Euronext, BATS and Direct Edge and NASDAQ run the biggest electronic markets, but bank’s and private equity firms’ ‘dark liquidity pools’, unseen by market data systems and the general public, also trade millions of shares each day.
The markets are run by quants - physicists, mathematicians and such - a new type of professional investor that earns money by computer-generating many tens of thousands of mostly small (a few hundred shares) trades each day. They often earn only a fraction of a cent per share, but their profits mount because of the huge number of trades they generate.
Electronic Exchanges are spreading. The London Stock Exchange’s Millenium system is now also in use by exchanges in Europe, Africa, Asia and the USA. The NYSE Euronext, in partnership with ATG, has announced plans to operate in Brazil as has Direct Edge; others have considered entering this market as well.
In the mid-1960’s, though, Wall Street had few, rather primitive computers and storage devices - and a lot of punch card processing equipment. Trading volumes grew from five million shares daily in 1965 to about twelve million per day in 1968. Traditional manual accounting, record keeping and stock transfer systems could not keep up.
Certificates for each and every trade had to be physically delivered and received. Paper shufflers, working overtime in brokerage firm back office sweatshops, struggled to hold the ‘Street’ together. There were not enough skilled veterans to go around. Long-time clerks became managers while barely trained ‘warm bodies’ took over the clerical functions. Systems that had served for decades crumbled as trading volumes grew and error rates climbed steeply.
Follow us on
Latest ICT EventMobile World Congress 2013
Mon Feb 25-Thu Feb 28
Download EXFO White Paper for FREE
• Africa and the Middle East
• Latin America
• North America
I was there and lived it.
The crisis was stopped - although the causes were not then resolved - by a long weekend of manual labor and a hastily rigged punch card system. Primitive, but it worked.
Clearly, changes were needed. In the years that followed the crisis, the industry - the New York Stock Exchange, its member firms, NASDAQ and others throughout the world - began rethinking their systems and operations and started using computers and communications for transaction processing and controls.
The exchanges led the way. New systems and services to do everything from speeding orders and automating trades, to clearing, compensation and custody were made possible by what was then state-of-the art technology. In those days just one megabyte of memory (ferrite core memory, there were no memory chips) cost several million dollars and only the NYSE (New York Stock Exchange) and NASA had them.
The advent of Depository Trust and net settlement systems transformed trade clearance and settlement from a system based upon the physical delivery of securities to one that depended only upon computerized bookkeeping entries to transfer ownership and effect payment. This alone, cut brokerage back office costs by 90 per cent in the first year in operation.
With back office problems mostly under control, and the ability to handle greater trading volumes guaranteed, the use of computer-based trading systems to speed and control the flow of orders to the world’s biggest exchange accelerated. The ability of exchanges to ‘electronically’ execute ever larger and more complex trades and order conditions grew apace.
NASDAQ, the first electronic securities exchange was founded in 1971 by the National Association of Securities Dealers to replace the OTC (over-the-counter trading market). NASDAQ, originally just a computerized listing of bid and ask quotations, did not begin to connect brokers to execute trades until 1987; it is now the second largest exchange.
On October 19, 1987, stock markets around the world crashed. The Dow Jones Index lost 508 points, a 22.61per cent decline - the biggest one-day loss in history. Fortunately, the market pulled back up rather quickly in most parts of the world.
‘Program Trading’, the computerized execution of trades for a basket of 15 or more securities - all at once when a specific set of conditions are met, was thought by many to have caused the this crash,. Others believe the underlying causes were a lack of liquidity, overvaluation, investor over-reaction and the like.
Because of this crash, many exchanges built automatic circuit breakers and other controls into their systems to halt or slow trading whenever prices dropped more than a given amount within a single trading session; they have rarely been used.
On May 6, 2010, the ‘Flash Crash’ occurred - the biggest and fastest drop in market history. The major indices dropped some 15 per cent that day - 7 per cent within a 15 minute period. About one trillion in market value was lost in minutes. Unbelievably, most of the losses were recovered in minutes. For the first time in history, an enormous market dive bounced back almost immediately.
The underlying cause and specific precipitating factors of this astounding market swing are a matter of debate. HFT, high-frequency trading, is one of the most frequently cited causes. Although many believe HFT wasn’t the underlying cause, it certainly facilitated the flash. People cannot trade that fast, but thousands of HFT transactions can be executed in less than a minute; trades are missed or made in nanoseconds.
The velocity of the trading process is limited only by the speed with which computers can run their algos and the time it takes to transmit the trade orders to the electronic exchange. Speed to market is a major competitive advantage. The HFT that gets a trade to the exchange a microsecond faster than a competitor wins.
Many exchanges now let HFTs co-locate the trading servers that run their trading algorithms in the exchange’s own computer center. The lengths of the cables connecting each of the HFT servers to the exchange’s ‘matching engine’ are rigorously equal, so only the HFT’s own computers and algorithms are responsible for any delays getting trades to market.
The newest links between the exchanges in New York and the Chicago Mercantile Exchange, where futures and derivatives are traded, beam microwaves by the shortest possible route. The best round trip times for data between exchanges in these cities varies from 14.5 milliseconds for the first fiber-optic network for this purpose to 8.5 milliseconds for the latest microwave system. The six millisecond microwave advantage would be a clear winner under all circumstances, were it not for interruptions caused by rain and losses due to certain atmospheric conditions.
HFTs need more than super fast access to markets and quotes. They need access to all the news that might affect the market - and they need it in machine digestible form. For several years, Dow Jones and others have been directly feeding more than 100 thousand predigested, ‘elementised’, news items to their clients’ trading computers. Human analysis is much too slow for a HFT. Quotes and news a few minutes old are ancient history to a high frequency trader.
Speed is an advantage, but it also puts High Frequency Traders themselves in mortal danger. On August 1, 2012, the Knight Capital Group lost US$440 million in about an hour when its trading algorithm ran out of control. Knight was the largest HFT house in the USA, trading more than US$ 20 billion per day with razor thin margins. Knight’s own stock plunged rapidly and it lost more than 75 per cent of its equity value. A group of investors bailed Knight out a few days later so it could remain in business.
The safety of electronic markets, and of the traders themselves, would be well served by an algorithm versus algorithm oversight and control system to cap the dangers, but allow markets to continue. The challenge is to develop ways to keep the market safe, not to reduce the risks or competitive advantages that are at the heart of any market.
Despite the apparent risks, most of the volume on all major stock exchanges throughout the world is now generated by high-frequency trading. At the NYSE close to 80 per cent of the business is generated by HFT. There have been a few scary moments, but in the end, the liquidity, the sheer volume of trading tends, or so it seems, to quickly correct major trading oscillations and damp the dangers.
HFT is an inevitable outcome of the computerized arms race between high-powered, high-stakes, financial market players. It can’t be avoided, but we need to understand it, control it and start thinking about how to control its inevitable successors. Yes, HFT can be dangerous, fortunately, though, it has tended to be reassuringly self-correcting.
|What can Connect-World offer you?
|Copyright © 2013 Connect-World. All rights reserved.|
|If you wish to cancel your subscription to this newsletter,|