Mikhail Kopytine gives an interview to Forex Hunter Blog

Wednesday, 18 November 2009 14:11

This interview appeared in the Forex Hunter Blog during that blog's brief but stormy life in November 2009. With Forex Hunter's permission, we reproduce it here.

FH: Please tell our readers a little bit about your background and how it affects your approach to trading?

MK: I am a physics experimentalist by education, trained to analyze and solve complex problems using a range of applied methods.

This requires, among other things, that a researcher be able to discover subtle statistical patterns in large amounts of data, patterns otherwise lost in the noise, as well as to write code for complicated research tasks, and to analyze output from research hardware. You have to be rigorous in your analysis, and be able to defend your ideas and conclusions.

On Technical Analysis

FH: What do you think about technical analysis? Can it be used to make money?

MK: Algorithmic trading which, we know, can make money does belong to the realm of technical (not fundamental) analysis. But I am skeptical of the broader forms of technical analysis where the question of randomness (or lack thereof) is often poorly addressed. One ought to be always mindful of the fact that a properly constructed random model can emulate many graphical signatures to which chartists ascribe significance.

FH: Are there any simple technical strategies that can be used to make money?

MK: It depends on what you mean by "simple." Everything that can be conceived by a mind is simple to its creator at the moment when it is created. But it may take tens of thousands of lines of code to turn it into a practically useful product and years of work to work out the details.

Initially, my algorithm was not only simple conceptually but a simple prototype could be coded in a matter of perhaps an hour. I recall I took a sine wave, fed it into the learning component and had it make predictions.

The catch: nothing inside the algorithm knew what a sine was, or that the input would be a sine wave. Yet the prediction on the basis of detectable redundancy alone (and sine is a highly redundant pattern) began to work instantly like a charm. At that point I understood that the idea was worth developing further. If it had not been simple initially, it would have been a lot more difficult to get started.

FH: Are there any misconceptions about trading that you would like to point out?

MK: The first thing that comes to mind is the concept of momentum -- people deal with the market as if it were a physically moving object having inertia. In reality, often this is very far from truth unless we are dealing with very special situations.

On Quantitative Trading

FH: Is there such thing as a trend, from a quantitative point of view?

MK: There are two parts to your question: 1) Is there a way to define what a "trend" would mean from a quantitative standpoint?

The answer to this question is yes. To me, a "trend" is a positive peak in the autocorrelation of returns with finite width (correlation length).

2) The next question is whether this sort of "trend" (defined mathematically) exists in actual markets.

Here, the answer depends on which market we're looking at, as well as the time frame (e.g., 1-minute time frame, 1-hour time frame, daily time frame), and the point in time in the market history. In the recent forex data (2002-present), trends as defined above are not visible on the 1-hour time frame. But I don't rule out the possibility that there are markets, time frames and time intervals where trends exist. The market for penny stocks may be one of them.

FH: How does quantitative trading make money? Can you describe one or two generic ideas?

MK: Here is a funny description borrowed from a book:

"They are sitting in the living room listening to computers, which are outfitted with neural net graphics packages and speakers, barking out commands to buy Treasury bonds and sell the Nikkei. Shirtless and stripped to their underwear in the afternoon heat, the men are quaffing large quantities of beer and wine while jockeying trades from Chicago to Singapore. DeNoyo's baby daughter crawls across the floor. His wife sits at a workstation writing code. The rest of the room is a bedlam of cheers and high fives as they watch the telltale signs that signal the approaching collapse of the Japanese bubble economy. The crash will wipe out two-thirds of the value of the Japanese stock market. DeNoyo is happily positioned on the short side of this momentous event. Every five minutes a computer screen flashes red and a synthesized voice resembling that of Darth Vader announces, "SELL FOUR JUNE NIKKEI." Out goes a call to a broker in Chicago, who has no idea he is dealing with a bunch of Texans sitting in their underwear."

(The Predictors: How a Band of Maverick Physicists Used Chaos Theory to Trade Their Way to a Fortune on Wall Street. T.A.Bass).

On a serious note, quantitative trading begins by pinpointing deviations of actual markets from the efficient market hypothesis, the latter being in some sense the case of maximum possible symmetry. Lack of symmetry is a distinctive feature of real life, including real-life markets. Symmetry is a signature of death. Once the asymmetry is quantified, it can be turned into the asymmetry of outcomes -- in your favor.

FH: What is a trading algorithm?

MK: A trading algorithm in my usage of the word is a decision-making tool enabling the operator to decide what, when, and with what allocation of capital needs to be bought or sold to maximize the utility of capital.

FH: Can you explain what the autocorrelation concept means and why it's important?

MK: Autocorrelation is a quantitative measure of the degree of predictability of a time series on the basis of itself. In brief, significant autocorrelation associated with finite time lags indicates that the game can be "cracked".

Observation of a significant autocorrelation at non-zero time lags indicates presence of statistical dependence between the points in time separated by that time lag -- not to be confused with causal dependence. In many cases, we don't care what the causal dependence is in order to be able to use autocorrelation.

FH: What are your thoughts on the concepts of mean reversion and momentum? Do these effects exist? If so, can a retail trader capture them to her benefit?

MK: I understand momentum to be essentially the same as "trend" in your other question. Both mean reversion and momentum are quantitative concepts testable with the autocorrelation technique.

In forex, stock indexes and LIBOR rates there is more evidence in favor of mean reversion (see the effect referred to as "bipolar disorder" in the research reports on the ForexAutomaton site) than there is in favor of momentum. In LIBOR, however, we see evidence of both mean reversion and momentum.

FH: Are there any other quant concepts that a non-quant trader should be aware of?

MK: It is easier to tell you what to ignore: I believe anything having the efficient market hypothesis at its foundation can be safely ignored by a practicing trader.

It is more important to develop what one might call "intellectual sophistication" or judgment in classifying various concepts. It is also important to keep an open mind. It is safe to say that most of the tools that work well "here and now" tend to be proprietary, carefully guarded ideas, not available to the wider public. But one can study what was used in the past.

The ForexAutomaton Project

FH: Why did you call your project "ForexAutomaton"?

We thought the word "automaton" had the right connotations within information theory and carried a highbrow academic flavor. For a project dealing with informational redundancies in the stream of financial quotes, information theory is the key. Since the goal is to automate decision making, the word "automaton" fits naturally.

FH: When did you start ForexAutomaton? What are the goals of the project? Will you be launching a trading signal service at some point?

MK: ForexAutomaton is now in its second year of operation, counting from April 2008. The first year of operation we were focused on detecting, quantifying, and classifying significant inefficiencies in historical forex data. Such research will continue in the future, and will continue to be documented on the ForexAutomaton web site.

The second year of operation, April 2009 through April 2010, is when we have been planning to launch a signal service, and it looks like a realistic goal at the moment, with an estimated launch date in early 2010.

At this early stage we expect our system to produce one trading signal per day. It will also run a model portfolio. A more limited version of the system will simply tell the user its directional forecast for the next close.

FH: Does ForexAutomaton provide strategy evaluation services?

MK: Yes, we do this on a case-by-case basis.

On Trading As a Business Venture

FH: Let's compare trading with other businesses which strongly rely on the laws of probability. Suppose one has $1 million in risk capital. Is trading better than, say, launching an online casino (assuming one complies with applicable laws), i.e. is trading a higher risk/higher return venture?

MK: As business literature teaches us, when choosing a business niche, one has to consider the entry barriers, buyer power, supplier power, threat of substitutes and degree of rivalry. If you are a guy from the street and you want to launch an online casino, chances are you will be fighting an uphill battle, confronting high rivalry, high buyer power and threat of substitutes from online trading businesses.

Speaking of substitutes, online trading can offer the same adrenaline effects to your clients without any stigma and perfectly legally, while making them feel good about themselves, being liquidity providers in the market. But, for the sake of a hypothetic argument, if you have unique ways of overcoming these difficulties, then an online casino may be for you.

In trading, the situation is more straightforward. The environment and especially the products are standardized to a very high degree, therefore your technique and your knowledge of yourself become the only factors you need to control to gain an edge. In most other businesses your individual qualities are only part of the picture. So if your unique advantages are in the area of self-reliant analytic work, self-knowledge and self-control, then trading is the way to go.

This is a strategic distinction, having nothing to do with risk/return and other efficient portfolio terminology -- most large capitals were created in ways which had nothing to do with efficient portfolios.

FH: How do you see the future of trading and markets in the XXI century?

MK: I see the XXI century as an age of crowds of kaleidoscopically diverse individuals, empowered and inter-connected by technology, interacting in more and more complex, conditioned and mediated ways -- and at the same time more and more differentiated within personal "virtual realities." I anticipate further inter-penetration between the concepts of information and capital.

Consider the price of commodities which is formed by trading contracts with a nominal value that, at any moment, far exceeds the value of physical commodities themselves. Obviously what is being traded is not the commodity itself but its idealized representation, an abstraction the essence of which is informational.

The other side of the same effect is the World Wide Web where ownership of information can literally become a source of passive income not unlike the more traditional bonds or land rent. Moreover, the interactive nature of the Web gives information another essential property of capital: the ability to reproduce itself.

The curious thing is -- from the point of view of information theory, there is no difference between a bit of information describing the color of a pixel of some porn star's bare butt in a movie transmitted over the Internet and a bit of information which is part of an ASCII symbol in a corporate communication with billions of dollars at stake or an important government communication -- as long as the information is efficiently encoded (that is, a good video codec is used in the case of a movie).

Information theory knows well how to handle information in the sense of efficiency, but not meaning. The video codec example, while a little extreme, lets you appreciate how the signal to noise ratio on the "information highway" in the sense of meaning will continue to deteriorate dramatically due to the growth of "noise" -- unless the theory learns to compute meaning (reconstruct generating paradigms behind sequences of data -- needless to say, such a discovery would revolutionize algorithmic trading as well).

Pending that, we are entering a market inefficiency hell (or paradise if you are an algorithmic trader) with swinging interconnected crowds of digital media users and weirdest distortions in the interests of those with ability to manipulate these crowds via concentrated input-side access to the media channels.

Thus, there is a promise of a highly predictable and volatile environment which is all one cares about in trading. To what extent market distortions can be fixed by algorithmic trading is, in my mind, an open question. What is clear is the fact that feeding on them, algorithmic trading does reduce them. In the environment shaped by these trends, trading, including algorithmic trading, will continue to be the means of differentiating the individuals, separating the minority of determined winners from the majority.

Bookmark with:

Deli.cio.us    Digg    reddit    Facebook    StumbleUpon    Newsvine