Skip to content

Trend-Following strategies (CTA style) to reduce downside risk and high correlation exposure.

Notifications You must be signed in to change notification settings

Djangu-algo/QARMII_project

 
 

Repository files navigation

QARMII_project

mat texTF

HEC Lausanne

Quantitative Asset and Risk Management II
Prof. Fabio Alessandrini
HEC Lausanne

The aim of this project is to create a CTA style cross asset trend-following strategy with two side objectives:

  1. Avoid portfolio crash during momentum reversal.
  2. Work in high correlation period.

Indeed, academic research has shown that trend-following strategies tend to under-perform on high correlation regimes between assets. Moreover, obviously, when market drop or rebound fast, the signal may take a while to change sign and therefore perform poorly.

Signals

We used many signals in order to asses their performance.

  1. Basic Signals
  • Momentum with varying length, espcially 90 and 252 days as well as the return of the month 9 to 12.

  • Moving Average Crossover with varying length

  1. Advanced Signals
  • Weighted normalized EWMA Crossover (based on this article from Baz. & al. 2015)

  • Singular Sprectal Analysis based signal

  • Support Vector Machine classification based signal

Weighting schemes

We combined these signals with three different weighting schemes :

  • Equally Weighted
  • Volatility Parity (inverse volatility, naive parity)
  • Risk parity
Constant Volatility

We also used a leverage to attain a constant volatility, allowing the strategy to be more easily compared.

Implementation

The implementation is performed on matlab, for each strategies we created a function that takes on the data and parameters, and compute the signals, weights and leverage at each rebalancing.

Author and Acknowledgment

  • Maxime Borel

  • Benjamin Souane

Thanks to Fabio Alessandrini for the help and to Kevin Sheppard for the amazing MFE toolbox.

About

Trend-Following strategies (CTA style) to reduce downside risk and high correlation exposure.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • MATLAB 95.1%
  • TeX 3.7%
  • C 1.2%