Directional Change

Einstein: "time has no independent existence apart from the order of events by which we measure it."
(The Universe and Dr. Einstein by Lincoln Barnett, Dover edition, 1957, 2005)


Directional Changes video on YouTube

Market prices are traditionally sampled in fixed time-intervals to form time series. Directional change (DC) is an alternative approach to record price movements. DC is data-driven: price changes dictate when a price is sampled and recorded. DC allows us to observe features in data that may not be observable in time series. Time series and DC-based summaries should complement each other: they allow us to see data from two different angles (like seeing things with two eyes instead of one). Our research agenda is to develop ways under DC to turn data into information, knowledge and applications.

Computational Finance and Economics Laboratory | PPT Presentation | Sponsor:   Olsen Ltd / Lykke


History is recorded by events, not end of year (or decade) snapshots

History is recorded by key events. For example, when we describe the development of currencies in the last century, we report when they untied with gold, when they floated against each other, etc. We don't take snapshots at fixed intervals: that is, we don't report the situations at the end of every year or every 10 years.

Yet, when we describe price movements in financial markets, we tend to use snapshots taken at fixed intervals. First we decide how often we sample the data. Then we take snapshots at the frequency that we have chosen. These snapshots form an interval-based summary. For example, end of day prices are frequently used.

Directional Changes Definition on YouTube

Directional Changes: data-driven summaries of price movements

Snapshots at fixed intervals are problematic. For example, if we record the end of day prices only, we wouldn't have noticed the flash crash on 6 May 2010. This motivated Richard Olsen to invent Directional Changes as a new way of summarising price movements.

A directional change is defined by a threshold that the observer cares about, e.g. 5%. Suppose we want to summarise the series shown here with 5% directional changes. We look for extreme points from which price dropped or rose by 5% or more.

Why are Directional Changes useful?

History is recorded by significant events. So should price movements. The advantage of using Directional Changes for price summaries is that it captures significant points in price movements. It would have captured the significant points during the 2010 flash crash.

This new concept provides traders with new perspectives to price movements (as demonstrated by Olsen Ltd in foreign exchange trading). This new concept has enabled researchers to discover new regularities in markets which cannot be captured by interval-based summaries. Such newly observed regularities give rise to new opportunities. As a new concept, directional changes open a rich research area waiting to be explored. Tsang 2021 argued that directional change is particularly useful for handling tick-to-tick data.

Sample DC summaries

Directional Changes Demo

Selected References

Projects based on Directional Change (in chronological order):

Researcher Project Remarks
Hvozdyk, Lyudmyla (PhD) Jumps in intraday data Advisor to Shengnan Li
Li, Shengnan DC-based Head and Shoulder trading, Relative Volatility between two markets, Jumps MSc by research followed by PhD, CCFEA, (2015.10-)
Alkhamees, Nora DC-based event identification PhD project, IADS, (2015.10-2019)
Golub, Anton DC Stylized facts, algorithmic trading PhD, flov technologies and Lykke, (2014?-)
KAMPOURIDIS, Michael (PhD) Algorithmic trading University of Essex
SUN, Jianyong (PhD) Overshoot analysis ex-CSEE, University of Essex
PANIANGTONG, Sanhanat Algorithmic trading MSc student (2014-15)
SERGUIEVA, Antoaneta (PhD) DC-profiling, high-frequency data UCL Financial Computing and Analytics
GAO, Jing DC-profiling Beihang University / Visiting Scholar (2014-2015)
TAO, Ran (PhD) DC-profiling: vocabulary and comparison PhD project, CCFEA, (2014.01-2018)
CHEN, Jun, James DC-profiling: Regime shift PhD project, CCFEA, (2015.01-2019.10.30)
MA, Shuai (Martin) DC-based market tracking and Nowcasting Visiting scholar, CCFEA, (2014.12-2015.02) / PhD project (2015.10.01-2022.03)
AO, Han (PhD) Forecasting DCs and algorithmic trading PhD project, CCFEA, (2012.10-2018)
BAKHACH, Amer (PhD) Forecasting DCs and algorithmic trading PhD project, CCFEA, (2014.01-2018)
YE, Alan Algorithmic trading PhD project, Greenwich University (2013.10-)
CHINTHALAPATI, Raju Venkata (PhD) Algorithmic trading Business School, Greenwich University
AL OUD, Monira (PhD) Agent-based stylized facts in the FX market PhD project, CSEE, (2009-2012)
MASRY, Shaimaa (PhD) Event-Based Microscopic Analysis of the FX Market PhD project, CCFEA, (2008-2013)
QI, Maggie (PhD) Risk measurement with high-frequency data PhD project, CCFEA, (2006[?]-2012)
TSANG, Edward (PhD) DC-profiling, forecasting and algorithmic trading CCFEA
OLSEN, Richard (PhD) Scaling laws Founder of OANDA and Olsen Ltd


This is part of the High-Frequency Finance Project. Some of the works decribed below were presented in the HFF Workshop 2016.

This page is maintained by: Edward Tsang; created: 2016.11.22; last update 2022.03.16