Machine-Designed Trading Algorithms Surpass Human Designs in Independent Testing

SAN JOSE, Calif. - April 21, 2014 - PRLog -- In controlled market simulation testing, machine designed trading strategies outperformed strategies developed by human Quantitative Analysts. Quantitative Analysts are highly paid Wall Street employees who develop strategies that are programmed into automated trading systems. Their algorithms are closely guarded intellectual property that can generate millions of dollars in trading.

In the 2014 Futures Truth Company report (the oldest trading strategy rating company), Trading System Lab’s (TSL) Machine-Designed Algorithms won the performance rating. In the verification method perfected by Futures Truth Inc., strategies are “run” against new, unseen market data and the performance of each one is captured. This is accomplished after the strategies are frozen, and no changes are allowed after this point. Poor strategies will decline, good strategies will rise. At incremental periods in the future, results of hundreds of strategies are consolidated and reported.

George Pruitt of Futures Truth commented: ”We have been running these tests for decades, in which we manage a ‘horse race’ where analysts can submit trading strategies. These strategies are measured over a fixed period against new market data. This is the first time in our research that a machine derived algorithm outperformed algorithms designed by analysts. This could signal a change in how traders develop their trading plans.”

“Machine-Designed Trading Algorithms” are trading algorithms created by a computer program which takes into account price, volatility, patterns, indicators, social media data, machine readable news and other data. A human generated equivalent can take weeks or months to perfect and often cannot take into account as many factors. The related academic field is sometimes called “machine learning”, referring to the ability of the program to use artificial intelligence to teach itself how to accomplish certain tasks, in this case to trade more efficiently.

The paradigm shift is that these strategies were designed by a Trading Specific Machine Learning Algorithm, not a human. Shown efficient at low- and medium-frequency trading strategy design, TSL now aims to show effectiveness on high frequency data. Using TSL and focusing on both speed and alpha, the machine can create and apply algorithms that consume order book dynamics as well as higher granularity information. This added dimensionality of Machine Learning at the high speeds available through TSL’s patented algorithm often produces better results than focusing on only one of these elements. Should regulatory oversight mandate changes to High Frequency Trading or order types, companies scrambling to create new algorithms at a rapid rate would benefit from this technology.

“Having an unlimited supply of Trading Algorithm Equity Curves offers a competitive advantage to anyone controlling a large amount of information”, noted Michael Barna, President of Trading System Lab. “Alpha mining will not get easier in the future and as we have seen lately, those who do not employ proper algorithms may not survive the coming market evolution.”

Trading Systems Lab is a privately held company in Silicon Valley with clients from Wall Street to Hong Kong.  TSL’s President, Michael L. Barna, is a former rocket and ramjet engineer in the defense sector and studied Astronautical and Aeronautical Engineering at Stanford.www.tradingsystemlab.com or info@tradingsystemlab.com