I. Introduction

For more than 10 years IntelliTrade's Principals have been developing novel and powerful methods of representing and analyzing market systems. Eight years ago we began using prototypes of our technology to support speculative futures trading. We started with speculative trading because that is what a major options market maker originally engaged Dr. Hoppe to do in 1989: develop trading applications of artificial intelligence. He had been working on such applications for five years prior to being engaged as a consultant. Between them, Dr. Hoppe and Mr. Wendling have devoted more than 25 man-years to developing the pattern representation and recognition technology IntelliTrade employs. Over the past two years we have been researching other applications of our pattern recognition technology, focusing on risk estimation and capital management, and it has become clear to us that they are better applications of our proprietary technology.

IntelliTrade's programs offer decision-making support to capital managers by providing reliable quantitative information about the behavior of selected markets in the web of international financial markets. IntelliTrade's programs do not replace human judgement, they support and inform it. The information our programs provide helps human managers do their job better.
 
 
II. Products
 
IntelliTrade's products fall into four classes, U.S. Fixed Income, U.S. and Japanese Equities, Currencies, and U.S. Asset Allocation. The three U.S. products are organized as a tree while the Japanese Equities Overlay Program and Currencies Program address problems in other domains.

The U.S. products can be integrated to support management of a common pool of capital as shown, or they can be executed independently of one another. (The red dots on the diagram indicate the IntelliTrade programs.)

III.  Tasks

To analyze a market to provide a basis for effective management one must perform three separate tasks:

A. Represent the Market System

IntelliTrade's program represent a market family as a set of MAPs which encode the web of relationships connecting the target market with the other members of the family. MAPs represent the linear and nonlinear relationships between markets through time. Our goal is to describe the web of markets in a way that allows us to look at a target market and assess how the reverberations in the web will affect it in the near-term future. Our programs map the patterns of relationships in the web of international financial markets in order to track disturbances as they impact a target market. Given the complexity of market systems and the complicated ways in which disturbances move through time and across markets, it is impossible for an unaided human to know all of the relationships and to accurately and objectively track complex patterns of disturbances as they move through the web. The human cognitive apparatus is simply not capable of dealing with that sort of system. IntelliTrade's programs have learned the relationships among markets and they have learned the characteristic ways that disturbances move through the web, and can support human decision-making.

B. Extract Useful Information

IntelliTrade extracts useful information from its MAPs by comparing the state of the market family today with a database of historical market family configurations, and selecting the sample of occasions most similar to the current state. Using that selected sample, the programs estimate current risk and calculate protective indexes. The sample of past similar configurations provides the basis for inferences about what the market is likely to do given the current state of the family of markets.

C. Use the Information

Organizations differ considerably in their needs and practices. Specifically how the several sorts of information provided by IntelliTrade's programs are used depends heavily on a particular client's goals, on its standard operating procedures and institutional limits and practices, and on other institution-specific variables. As a consequence we cannot specify a single generic or standard application. We have developed examples of some ways the information can be applied, but clients must adapt the information to their own circumstances. IntelliTrade supplies over four years of historical data to aid organizations in designing specific applications. The programs do not make mechanical decisions. Rather, they provide information to support decision-making. IntelliTrade's developers are available to assist clients in developing custom applications based on the information provided by IntelliTrade's programs.
 
IV. Foundations

A. Related Markets Form Webs, or Families.

This diagram represents a family of markets. Each node in the diagram represents a market. The center node is the target market being analyzed. The lines represent relationships between markets. While there are a large number of pair-wise relationships connecting all the markets (some are shown as straight lines), our programs represent only the relationships between the peripheral markets and the central market under analysis, represented by the heavier lines.

We regard financial markets as forming a complex web, with perturbations flowing through the web from market to market in time. Suppose you are walking in the woods and find a spider web across your path. If you pluck one strand of the web, the whole web will begin to vibrate. The original disturbance of a single strand is transmitted across the web in complex patterns of vibrations. The web will continue to reverberate long after the original stimulus is gone. Market systems are similar. A disturbance in one market is transmitted across the web of inter-related markets and the system reverberates in complex patterns for some time after the original disturbance has ceased. The market web is in a constant state of vibration.

We do not represent the relationship between the target market and all possible other markets. That is a computationally intractable problem. We have done a great deal of research to narrow down the relationships and nodes that must be included in order to give a reliable representation of the target market's "family." For example, in the IntelliTrac Fixed Income 30-year U.S. Treasury Bond program we use just six peripheral nodes -- the 2-year, 5-year, and 10-year U.S. Treasury notes; two foreign interest rates, the German Bund and the British Gilt; and the Dollar/Yen exchange rate. Those six peripheral nodes, along with the 30-year bond itself, are sufficient to capture the primary characteristics of the bond family's behavior. One could add any number of markets to the periphery of the web, but we have found that more nodes add little to the description of the long bond's behavior and their addition can actually hurt the generality of the representation through time by presenting the opportunity for spurious relationships to be learned by the programs.

Our goal is to describe the web of markets in a way that allows us to look at a single target market and assess how the reverberations in the web will affect it in the near-term future. IntelliTrade's programs have learned the relationships among markets and they have learned the characteristic ways that disturbances move through the web.

B. Prices Carry Information About Relative Values in Families.

From economics we adopt the view that the primary function of prices in free and liquid markets is to carry information about perceived relative value. Changing prices signal changing perceptions of relative value. The changing perceived values influence the movement of money between markets which in turn recursively affects prices. By mapping market families and tracking price changes within a family of related markets through time we can infer how perceived values are changing in that family and how money is likely to flow between members of the family in the near-term future. We therefore use daily open, high, low, and closing market prices as the raw inputs to our programs. The time horizon we attempt to foresee ranges from one to thirty business days.

C. Market Systems Are Not Random.

We know that financial markets are non-random, non-normal, and non-linear. That is, the way prices move through time and the distribution of price changes cannot be adequately described by standard linear statistical methods. For example, in a paper to be published soon in a major risk management journal, Dr. Hoppe has shown that portfolio Value at Risk estimates based on standard statistical methods are seriously flawed. We have developed ways of representing the patterns of a market family's behavior that are independent of the assumptions of randomness, linearity, and normality and that do not depend on standard linear statistics.

D. Those Who Do Not Know History Are Doomed To Repeat It.

We know that markets -- or really, market participants -- have memories. The way a family of markets is behaving today is similar to the way it behaved in the past in similar circumstances. The received statistical wisdom is that markets follow random walks and therefore price histories are useless. That assertion is false. It is based on linear statistical methods that are incapable of detecting all of the relationships within and between markets. As people who manage real money in the real world know, market behavior is very complex but it is far from random. IntelliTrade's programs search the past to find occasions when circumstances were most similar to today, and they use the behavior of that sample of similar past occasions as a guide to assessing the likely behavior of the market in the near-term future. We use selected historical data to support quantitative inferences about the behavior of the market system. Historical information informs current evaluations.
 

V. Institutional Functions Supported
 
A. Risk Management

Information provided by IntelliTrade's programs supports the management of capital that is exposed to interest rate risk, foreign exchange rate risk, and equities market risk. For example, the Probable Path Analysis in the Fixed Income package provides a daily probabilistic description of the near-term future path of four U.S. interest rates based on selected samples of similar past occasions. The Hedge and Maturity Recommendations provide tested support for managing interest rate risk through time. We have developed examples of how interest rate risk to capital can be reduced using IntelliTrade's analyses and examples of how the several risk management decision support programs can be combined to support the management of diversified portfolios of domestic bonds and equities. We assist clients in devising their own applications of the information our programs provide.

B. Transaction Planning

The information provided by IntelliTrade's programs supports transaction planning. Knowing the probable near-term movements of markets, one can plan transactions to take advantage of likely favorable terms or to avoid likely unfavorable circumstances. To suggest one example, the Fixed Income program has correctly anticipated (in real time) the decisions made in the six most recent Federal Reserve Open Market Committee meetings, the anticipatory signals beginning several weeks in advance of the meetings. Correctly anticipating the decisions made at those meetings allowed planning fixed income transactions appropriately and profitably around the meetings. Knowing some weeks in advance that the probability of Fed action to raise rates at its March meeting was high, a prospective bond issuer would have been well advised to time the issuance to occur before the meeting or to have appropriate hedges in place before the meeting.

C. Capital Management

The information that IntelliTrade's several programs provide can be combined to manage diversified portfolios of equities and bonds. For example, The Asset Allocation Model can provide optimal allocations across equities, bonds, and cash, and the Fixed Income and Equities Overlay programs can guide the modulation of positions within those allocations to provide increased absolute return and risk-adjusted return.