Technology-based vocations constantly evolve. As tools improve, how we do our work changes. Most of the time, this is for the better. Sometimes, we take a step back. But one thing is certain: Change happens. This certainly has been the case in trading, where technology has redefined nearly everything. In this series, we will explore the birth and growth of technology in trading and, it is hoped, learn lessons that will help us better navigate what the future brings.
The concept of applying technical analysis to the markets has existed since the early 1900s in the Western hemisphere. Hand-drawn charts were used, as well as manual calculations of price averages. As time went by, companies sold chart book subscriptions, printed on Fridays and delivered on Mondays.
For many traders, though, this was not good enough. Contemporaries such as Larry Williams and Perry Kaufman began exploring using mainframes to calculate technical indicators at what seemed like lightning speed.
Interviewed for this article, Williams recalled backtesting strategies using punch cards in 1966. Kaufman said he worked with aerospace engineers from Rand Corp. and TRW during the early to mid-1960s, using moving-average-based systems to trade equities. A few years later, in 1970, Richard Donchian first published results of his four-week rule, optimized using powerful machines.
The backtesting and hard coding of trading rules tracked the growth of computer accessibility in corporate and educational settings. The main player was the PDP-8, introduced by Digital Equipment Corp. This was the first commercially successful minicomputer, selling for $18,000, one-fifth the price of a small IBM 360 mainframe. The speed, small size and reasonable cost enabled the PDP-8 to go into thousands of manufacturing plants, small businesses and scientific laboratories.
Engineers began exploring the functionality of these machines, and it wasn’t long before many recognized the applicability to trading. For the first time, the tools were driving the trading, and spawning new traders in the process. Technology was so powerful that not traders, but engineers and technicians who had access to cutting-edge mainframes, were shaping how market analysis would evolve.
In 1966 Hewlett-Packard entered the general-purpose computer market with its HP-2115, offering computational power formerly found only in much larger computers. It supported a variety of languages, among them BASIC, ALGOL and FORTRAN. Now computers could be programmed and rules could be written in these languages and tested. Still, computers cost $15,000-$20,000 (more than $100,000 in 2013 dollars). Despite the limited availability, those with access still developed a taste for analysis, working after hours on experiments.
While professionals were applying the power of mainframes, computerized technical analysis for the general public was ushered in by the hand-held calculator. On June 25, 1974, the U.S. Patent Office granted Texas Instruments patent No. 3,819,921 for a “miniature electronic calculator.” By 1975, the technology was widely used by students as simple, inexpensive four-function calculators. Calculators improved rapidly, and by 1977 the TI-30 scientific calculator with LED display began selling for less than $20.
During this same time, the personal computer came to market. The period of 1976-78 saw the birth of the Apple I, II, the TRS-80 and Commodore Pet computers, as well as VisiCalc, the first commercial spreadsheet program that sold more than 100,000 copies within the first year of its release.
Also by 1978, the 5.25-inch floppy drive began to spread. In 1980, Seagate Technology created the first hard disk drive for microcomputers, the ST506. The disk held five megabytes of data, five times as much as a standard floppy disk, and it fit in the space of a floppy disk drive. These first early hard disks were expensive, about $5,000. In 1981, the IBM PC machine that would change the world was introduced. MS-DOS and the 3.5-inch drive followed. In 1982 Lotus 1-2-3 was released and VisiCalc, which did not offer an IBM PC version, faded away. IBM clones were launched by Compaq in 1983, manufacturing costs plummeted and IBM’s architecture became the industry standard.
Meanwhile, retail-oriented technical analysis software was being developed for the Apple II. Trader Tim Slater realized that the Apple II could access large amounts of data capably, calculate technical indicators quickly and display them on its 10-inch monochrome screen. Loyola University Computer Science Professor James Schmit was engaged for his BASIC programming skills, and they planned, programmed and completed the first commercial CompuTrac, version 1.0, in July 1977.
To finance the development of CompuTrac, Slater commercialized the first version with three fellow market technicians who had been following the product’s early development. The Technical Analysis Group (TAG) was incorporated with Tim Slater as its founder and Richard L. Redmont, Jim Sibbett and Walter Bressert as original members. Shares were floated and issued.
CompuTrac was the first to implement and popularize many of the technical indicators used today. An example is the stochastic oscillator, which was programmed using the work of Ralph Dystant and George Lane with the indicator’s lines named “%K” and “%D.” However, Slater wanted a more marketable name; hence, the study became “stochastic,” and the term has persisted since. Another example is the “moving average convergence/divergence,” or MACD, study by Gerald Appel.
This rapid birth and popularity of computers, along with their affordability to private traders, led to the development of personal computer backtesting software. Two traders created the first commercially available software around 1982-83. Working independently, Louis Mendelsohn and Robert Pardo became the forefathers of modern day testing software.
One evening after work, hospital administrator Lou Mendelsohn stopped by a Radio Shack store to check out the “microcomputer” that he had just read about in The Wall Street Journal. It was a Model T Ford compared to what’s available today, but it was much more accessible than the mainframe IBM 360 that Mendelsohn was familiar with from his undergraduate days at Carnegie Mellon University in the 1960s. Then, using punch cards or a teletype machine, Mendelsohn had to submit his computer programs for processing and wait at least 12 hours to see if they ran successfully or had syntax errors. With a microcomputer, debugging took minutes instead of days.
“Right after I first saw a demonstration of a personal computer, I realized that this new technology would revolutionize technical analysis and the financial markets forever, not to mention every other sphere of human activity,” Mendelsohn says.
Mendelsohn set his goals high. He says he set out to develop the most powerful commodity trading software that could be programmed at that time. To speed his development effort, he attached a five megabyte external hard disk (which he still displays in his office today) to a single Apple II+ computer. Eventually, to help compile source code into object code during debugging, he had 11 Apple II+ computers on a local area network attached to a 40-megabyte Corvus hard disk.
In 1979, Mendelsohn started his own trading software company, Market Technologies (originally called Investment Growth Corp.). His goal was to gain a competitive edge over other traders using his software, as well as make it available to other individual traders becoming exposed to computerized technical analysis.
Shortly after his first son, Lane, was born in 1980 and despite admonitions from his wife’s father, Mendelsohn left his career as a hospital administrator to devote full time to his passion as a commodity trader and trading software developer. This life-changing decision led to the 1983 introduction of ProfitTaker Futures Trading Software, the first commercially available strategy backtesting and optimization software for the personal computer in the financial industry (see “Anyone welcome,” right).
In February 1983, Mendelsohn submitted an article to Commodities (later that year renamed Futures) magazine addressing the unique intersection of computing technologies and technical analysis (see “Picking software programs: Know their limitations,” May 1983, Commodities). By then, Commodities had published a number of articles about technical analysis, but mostly covering visible chart patterns and trading techniques, not how individual traders might use a personal computer to develop, test and optimize trading strategies.
It was clear by then that the personal computer had the potential to change the face of financial market analysis and trading. Mendelsohn’s article demonstrated that he knew and understood the subtleties and nuances of commodity trading, which are unique and difficult for many traders to grasp, much less write about and program into computer software.
Mendelsohn’s breakout article was followed by two subsequent ones in which he used ProfitTaker to introduce strategy backtesting to commodity traders (see “History tester important factor in software selection,” July 1983, and “Execution timing critical factor in system performance,” December 1983).
Given the limited computing power of PCs in 1983 compared to what they can do today, Mendelsohn’s first history tester had the ability to link actual contracts and test trading strategies under simulated, yet real-time, conditions; optimize various technical indicators and execution times; and handle rollovers of expiring contacts (see “Rolling along,” below).
Mendelsohn says that he grappled withmany conceptual issues, all of which affected the integrity of the test results. For instance, should the rollover date be hard-coded, such as a specific day of the month, or should it be an optimizable variable based on the spread difference between prices of the two contract months? Also, if the history tester is in a position in the expiring contract month on rollover day, should it close out that position and enter the same position on the following contract month, or should it have the intelligence to test the following contract month data leading up to the rollover day to see if the trading model would have been in a position in that contract? If not, should the history tester close out the expiring month position and stand aside until a new signal in the new contract is triggered?
Another challenge was locked-limit conditions. Mendelshon felt that the tester should check for these to prevent phantom profitable trades that could not have been executed realistically under real-time trading conditions. This issue was challenging to him because of the complexity of the locked-limit phenomenon. Each commodity had different limit amounts that could expand on subsequent days. Additionally, the exchanges changed the limits from one time period to another depending on market conditions. These data were not stored along with the open, high, low, close, volume and open interest, so building them into the programming was a chore.
ProfitTaker had several other unique capabilities. Each daily update displayed the current position status of each market being tracked. Additionally, it displayed what threshold closing prices on the next day’s close would trigger new signals to get into or out of positions rather than waiting until the following day’s open. ProfitTaker’s trading signals were long and short “sensitivity bands” that surrounded various moving averages. The bands could be optimized individually for both the long and short side of each commodity.
Other commodity traders followed in Mendelsohn’s footsteps (including several of his own customers). This paved the way for today’s huge technical analysis and trading software industry. One of the most successful of those companies was Omega Research, founded in 1982. Omega Research introduced its backtesting software with System Writer in 1989 followed by TradeStation in 1991. Bill and Ralph Cruz developed Easy Language for non-techies to program their own trading strategies.
ProfitTaker’s history tester (known as ProfitAnalyst) was reasonably fast when run on an external hard disk attached to an Apple computer or run on an IBM-XT with an internal hard disk. The trader simply configured the history tester on which commodities to test, what specific contract months (with rollovers) to include and what range and increments of model parameters to test for each commodity. He also was able to preset several trading performance criteria so that only those models that met all of the criteria thresholds either would be saved to a disk file or printed for further evaluation (see “Trade performance,” below).
The technical analysis premise behind ProfitTaker was based on Mendelsohn’s research into work by Frank Hochheimer, manager of technical analysis and computer applications for the Merrill Lynch’s commodity division. He also was an avid student of Kaufman’s two books, “Commodity Trading Systems and Methods” (1978) and “Technical Analysis in Commodities” (1980).
Like the computers he used from the time he was a one-man army to when his software company had grown to more than 50 employees and thousands of customers, Mendelsohn also evolved. He outgrew ProfitTaker in the late 1980s when he took notice of the globalization of the financial markets and redirected his research into forecasting them using intermarket analysis.
Floor to screen
Robert Pardo left Salomon Brothers in the fall of 1980 and entered a partnership with a floor trader who funded his research. The purpose was to develop trading floor buy/sell signal software using scalping numbers.
Pardo realized the advantage of having a computer and a database that he could use to simulate the actual trading of the method over its history and see exactly what the strategy would have produced. He then spent the majority of his time coding and researching various trading ideas. To do that, Pardo developed a trading simulator with optimization capabilities. The interface was crude, but it was fast and capable and could use new data vendors such as CSI.
By the end of 1982, Pardo was looking for a way to earn income so he could continue his research. He started by developing a number of trading utilities and the first version of Chartist. He also marketed a family of products as CAT (Commodity Analysts Toolkit). Chartist was a quick and rather complete set of charting, support and resistance studies and indicators. At that time, most of the competition did not offer any support and resistance studies, so Pardo’s work offered an advantage. He claimed that all the tools were effective, but rather complex for the average trader.
Pardo then met Chicago Board Options Exchange trader Norman Winski around March 1983. Winski suggested that Pardo attend a popular trading show, where Pardo sold more than $3,000 of product. A new business was born.
Pardo also did some consulting work during those early years and a young trader approached him about coding a trading method. This method became the trading core of Swing Trader, which was introduced in late 1983 (see “Swinging for profits,” right). Later, he released Advanced Chartist bundled with Swing Trader and sold it for $1,395. Support is an issue for small companies selling backtesting software, and once Pardo started doing more trading, selling the software was no longer an effective use of his time.
Early backtesting software developed in tandem with the technology of the time. It was a direct result of the availability of, first, the calculator and then personal computers. Fueled by the dream to understand the markets, engineers and early software developers turned to trading. Ultimately, however, their passions came full circle as they realized that marketing their analysis tools as commercial products was a viable supplement to personal trading activity.
The evolution of trading software did not stop there, though. Soon, technical innovations that allowed the automation of walk-forward testing, portfolio-based analysis and real-time trading would once again change how traders would interact with the markets. In our next installment, we’ll explore these developments and examine the lessons learned.
Murray A. Ruggiero Jr. is the author of “Cybernetic Trading Strategies” (John Wiley & Sons). E-mail him at firstname.lastname@example.org.