When our team first assembled and began this journey in 2009, we understood that customers require investment solutions with low volatility, and reliable, sustainable returns, expressed in high-uncorrelated alpha. Our goal was to provide transparent investment solutions to our customers, to meet these demands.
We knew that a new technological approach, incorporating latest state-of the art modeling and field-related research, would have to be developed, in order to have a market model, where the reflexive nature of markets and their players could be considered. Because only through this approach would investors finally be able to find peace in their trading decisions and stability in their assets. Let me go into a few thoughts on current approaches in the market, and what makes TwoStone´s approach so unique.
Many, if not most sophisticated investors share one common point of view of the market. There are numerous players who continuously react to rapidly propagating data. While information symmetry is usually quickly attained, “thought patterns” differ, as do reactions amongst players. This leads to complex market patterns, which are perceived as being too difficult to predict.
Reaching a state of equilibrium cannot be achieved under the conjectures above, as the system is in a constant state of information and reaction-driven flux. Some may argue that from a practical perspective, the concept itself is poorly defined, due to pluralism of paradigms among players.
In such a setting, interdependencies between different observables are dynamic, and the concept of “misalignment” becomes vague (apart from when referring to a specific point in time).
Taking into account that each “thought-pattern” can be represented by a well-defined model, the market itself becomes a dynamic, volume-dependent, “superposition” of these models. Another way to look at this system is to imagine it as being a dynamic force field governed by complex physical rules of motion. While the convergence point at each moment can be hypothesized, trading depends on the dynamics of the system, which add to the complexity.
With the nature of stand-alone technical tools being simple and at times even multivariate (as underlying variables may be tricky and rigorous to define), such tools have been abandoned by most experienced practitioners. When deployed, comprehensive elements, tools, and models first have to be added, in order to achieve measurable value.
More sophisticated proprietary forecasting tools utilize adaptive, multivariate, non-linear engines. These tools are becoming more widely used by the industry. The ideas and methods are borrowed from a wide array of disciplines, such as pattern recognition, heuristics, and self-learning algorithms, which are all fed with a wide array of observable data. Mathematically, these tools are similar to adaptive, model-switching tools. However, despite their sophistication, the “predictive power” of many of these tools is insufficient in allowing consistent direct translation of forecasts into trading strategies.
Too large for consistent direct application, both magnitude and time-domain errors rapidly increase significance levels. Typically, in high frequency scenarios, errors are small; yet friction kills gains. Lowering the frequency increases exposure to larger model errors, once again attenuating and squelching gains.
Another important fact to acknowledge is that current market volatility is greatly influenced by both psychological factors and technology. Investor expectations, speculation, portfolio stop-losses, derivatives, CDO's, and so on, all lead to technical chain effects.
Despite the sophistication of today's proprietary forecasting tools, predominant trading strategies still fail to yield sustainable returns. We believe that this is due to the very limited and present nature of financial modeling and econometrics.
In order to properly predict trends, it is imperative to develop truly predictive models, which in turn requires a realistic view of financial markets, and a concise interpretation of the emergence of financial time series and their stochastic properties.
Investing should be dull. It shouldn't be exciting. Investing should be more like watching paint dry or watching grass grow. If you want excitement, take $800 and go to Las Vegas. Paul Samuelson, Nobel Memorial Prize in Economic Sciences, 1970
In 2009 we identified the missing component, a reflexive market model, which translates the psychology and self-updating behavior of heterogeneous actors into a logical grid of model structures. This model space is continuously fed and informed by observable data, best describing the human-driven nature of markets.
Based on a dynamic projection into that model space, our forecasting approach proves to be superior and robust, regardless of structural shifts in the market. Modeling stochastic processes as multivariate model structures allows our technological approach to track interdependencies and misalignments between an individual time series.
Characteristic quantitative trading approaches, based on stop, start, and exit prices do not tell the full story, and in our view, will not yield sustainable performance in the long run.
Financial time series may be interpreted as complex stochastic processes with trends. Any financial time series can be forecasted for which the law of large numbers may be justifiably applied. Rather than providing exit, stop, and starting price values, we dynamically calculate price expectations and structural shifts in market prices. We remain ahead of the univariate or technical trend forecasts, by combining our strengths of low volatility and low frequency strategies to meet our customers' demands in having reliable, sustainable returns.
Our approach represents a major milestone in the evolution of trading strategies, and we invite our customers to become part of it.