Regular readers will recall the series of four columns from May 27th to June 17th that discussed the enormous profit potential in directing expert traders (now I prefer the term "talented") to bubbles and crashes. Over the summer our Computational Finance Center team at Tec de Monterrey tested Neural Stock® as a talented trader, applying it to a selection of bubbles and crashes from around the world. Neural Stock made money before transactions costs in this experiment; with modest simulated transactions costs, it only broke even. Since I staked part of my reputation on this application earning 5,000% annualized returns on its simulated capital, these overall results disappointed me (especially after the excitement when first day’s 1.85% average return for a 90-stock- portfolio -- 4,000% annualized -- had made my goal look achievable).

Nonetheless, the overall test results do not reject the hypothesis that a team of talented human traders could, with proper AI and capital support, easily earn 50 times their capital in a year. First, the AI platform we tested used neither high-frequency data nor non-linear optimization for its neural net inputs. The former feature should enhance day trading significantly, the latter should improve performance at all trading frequencies. More importantly, we did not put human traders into the loop, and teams of human traders with AI support will likely outperform either talented humans or talented AI systems alone.

Our shoestring budget kept us from taking these obvious and important steps. While Tec generously provided access to basic hardware, and CFC patrons Reuters® and Neural Stock generously provided data and software, our CFC team had no funding to support specialized software acquisition or trader screening efforts. Now that Neural Stock’s developer is far along in designing and testing both high-frequency data and non- linear input optimization features for inclusion in newer versions, a test with that newer version would provide more definitive results. However, since Tec no longer supports my participation in this line of research, to continue I must find new partners with the vision to fund this research in return for the opportunity to apply it.

Now, anyone can read The Economist’s survey and see that developers around the world are working to improve talented AI traders. Anyone can read Business Week’s report and see that reasonably-talented human day traders can earn from one to four or more times their capital in a year -- without trading inside the exchanges. I’m almost certainly not the first person to visualize the convergence of this advancing financial technology with the rise of massive off-exchange day trading, but I believe I’m the first person to see that this convergence will inevitably create a massive increase in trading volume, volatility, and returns as day traders learn to pile into all into the world’s bubbles and crashes, and I’m virtually certain I’m the first to write about the extraordinary consequences of that triple convergence of technology, trading, and volatility.

As both a researcher and a responsible citizen, I’ve always hoped to be the first person to harness that triple convergence and demonstrate its power first to my patrons and then to the world’s exchange regulators. Until I can overcome the barriers of about US$300,000 in research funding and about US$2,000,000 in capital, though, I won’t be able to build the facility that just might multiply that capital fifty times a year by making bigger bubbles and deeper crashes all around the world. As time passes, I fear someone else will do it first -- perhaps a reader of this column who is learning about this triple convergence -- but with the exclusive intention of accumulating as much trading profit as possible, rather than also alerting the world’s market regulators to a coming danger.

That fear brings us to the third recent event -- excess volatility in the peso/dollar exchange rate. When the triple convergence holds sway, the start of any localized excess volatility will lead to far bigger movements than we see today. When talented traders around the world, emboldened by their talented AI support and their instant liquidity access, jump right quick into any emergent trend, you can bet the dollar will rise far higher far faster -- the peso will fall far lower far faster -- than what we just saw in the last few weeks. That prospect should frighten market regulators even more than the public.

Readers with questions or comments for Dr. White can call 011(525)595-6045, fax 011(525)683-5874, or email white@profmexis.sar.net

Return to Mark White's Home Page

This page hosted by Get your own Free Home Page