Machine-Learned Ranking Algorithms for Implied Volatility Prediction

ABOUT US : Collapsed
INSIGHTS : Collapsed

Machine-Learned Ranking Algorithms for Implied Volatility Prediction

Download PDF
Caleb Weaver Ben Perryman Jun Jiang Alex Chien Erez Meoded Al Kirk
JUL 09, 2020

We recommend

Surface Classification for Direction Predictions of the Foreign Exchange Implied Volatility Surface

Functional PCA for Implied Volatility Surface Prediction

Predictions of the value of the implied volatility of an FX pair are commonly used as inputs to pricing models in an FX option trading strategy. Depending on the trading strategy, however, a metric that quantifies implied volatility relatively across currencies can be of supplemental value or of more interest than level prediction for a single currency.

  • Predictions of the relative magnitude of movement among the implied volatilities of FX pairs using pairwise ranking algorithms typically used for collaborative filtering and search algorithms consistently outperform predictions derived from standard implied volatility forecasts.
  • Rank modeling shows that PCA components derived from equity and commodity indicators are the key drivers for both short and long-term forecasts and the change in implied volatility surfaces are of primary importance in short-term forecasts.
  • The ensembling of independent forecasts of level, direction, and relative magnitude of implied volatility can be used in establishing forecast confidence and informing a trading strategy.