Tao Ran completed his PhD studies at the Centre for Computational Finance & Economic Agents, University of Essex in 2014-2018. He was part of the Directional Changes Project. In his research, he attempted to extract information from data.
Traditionally, researchers use Time Series to summarize and analyze price movements. Directional Change (DC) offers us an alternative way to record price movements. Instead of sampling data at fixed intervals (as it is done in time series), DC is data-driven: data is sampled at points when prices change direction (from going up to going down, or vice versa). Exactly what constitutes to a directional change can be found in [Tsang 2010] Briefly, when prices are going down (up), if price has gone up (down) from the lowest (highest) point by x%, then we call that a directional change. Different people will consider different x significant.
It is worth contrasting DC analysis with time series analysis: In time series analysis, the researcher determines how often data is sampled. in other words, the researcher determines the scale of the x-axis. In DC-based analysis, the researcher determines how big a change is considered significant. In other words, the researcher determines the scale of the y-axis. We believe that time series analysis and DC-based analysis look at data from two different angles; they should complement each other.
To enable researchers to analyze market dynamics in DCs, we need a vocabulary to describe observations. In this project, we invent indicators for describing DC-based market summaries. These indicators make up our vocabulary describe observations. Such descriptions help us to establish DC profiles to capture characteristics of markets. We believe that DC-based profiles complement observations in time series analysis. For example, the two analyses would provide different indicators for measuring volatility.
With the indicators established, we embark to invent metrics for comparing different DC-based market profiles. We intend to use these metrics to contrast different markets, or differences time periods of the same market. The aim is to provide insight to markets which may not be observable in time series analysis.