As our marketing efforts gradually shift from focusing on individuals to institutions, we have been asked recently, more than once, to provide a theoretical framework for our investment philosophy and trading approach. Although our trading results continue to validate our strategy, we were more than happy to take on this challenge, go back to review the genesis of our ideas from over a decade ago and review why our methodology still stands to reason.
The three pillars behind our approach to research and trading are the following:
- A given financial market reflects an underlying, aggregate psychological state of its market participants, which is bi-polar in nature alternating between extremes in bullish and bearish sentiment states. This belief is supported by the growing field of behavioral finance, which stands at odds with the efficient market hypothesis (EMH) school of thought. Behavioralists argue that market participants are subject to a number of biases, which can affect their investment behaviors. We believe that trading models can be designed to exploit these tendencies.
- The aggregate psychological state of a given market can be tracked via crowd sampling1. The advent of the Internet and the ensuing development of the field of computational linguistics have provided the basis for us to track the crowd’s sentiment with respect to the US stock and bond markets over time. As such, we have now been tracking the daily sentiment of the US equity market for 13 years and over a decade for the US bond market. Our research has shown that investor sentiment evolutions are quite distinct and, not surprisingly, tend to repeat over time. Additionally, the financial press represents a very compact and uniform linguistic space, which exhibits a number of specialized features, which are well-suited to machine learning methods.
- The crowd itself tends to both under-react to “real” news and over-react to “false” news over time2. We use our sentiment model to gauge such reactions and compare them to similar ones in the past. Our trading models then employ machine learning to condition relative value indicators on our proprietary sentiment dataset. Our four trading sub-strategies, each with its own particular behavioral aspects, generate trading signals with respect to both the expected direction of the market and, perhaps more importantly, changes in its associated volatility across realized, implied and expected moments.
In summary, our trading approach relies on both computational linguistics to track the psychological state of a financial market and behavioral finance methodologies as a means to take advantage of trends and reversals in aggregate crowd sentiment.
Madison Park Capital Partners (“MPCP”) is an opportunistic trading firm, which seeks to capitalize on short-term market dislocations within S&P 500 & VIX derivatives complexes. The firm’s philosophy is rooted in the behavioral finance axiom that investors tend to both over-react to “false news” and under-react to “real news”. MPCP’s trading leverages its proprietary database and extensive research on 13 years of equity and bond market sentiment. MPCP derives sentiment using data mining and computational linguistics to parse the financial press by machine reading roughly 2,500+ articles daily and extracting sentiment from those that are deemed relevant. MPCP has identified a number of “sentiment evolution patterns”, which the market cycles through over time. Using machine learning it generates trading signals conditioned on these patterns and exploits them across four sub-strategies. Trading is driven systematically with a discretionary overlay to optimize relative value relationships.
1Surowiecki, James. The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisom Shapes Business, Economies, Societies and Nations. Surowiecki argues that a “wise” crowd of individuals can produce a better joint solution than any single individual or group of “experts”. Four conditions must be met for the crowd to be considered a “wise one”: diversity of opinion, independence, decentralization and aggregation.
2Please see, Boudoukh J., Feldman R., Kogan S., and Richardson M., Which News Moves Stock Prices? A Textual Analysis?, 2013, NBER Working Paper No. 18725, January 2013.