Date: Wednesday 6 March 2019
Time: 12.00pm - 1.30pm
This seminar was delivered by Henri Waelbroeck, Vice President and Director of Research for Portfolio Management & Trading solutions at FactSet.
Volatility forecasting plays a key role in risk and options pricing. There has been a lot of work on the mean reversion properties of volatility, including GARCH, rough volatility models etc. Such models are limited to time-series data and therefore do not provide any insight on innovations. Earnings announcements for example represent the majority of annual volatility in most stocks. The volatility of Pharmaceutical stocks is dominated by news on the drug approval process, and so on.
Here, we review how machine learning can be used to refine a fair market price of single-stock options and how the options montage can be mapped to various jump models. We review some examples: a fit of the trinomial jump process to the 2018 mid-term elections in the US, the binomial model to fit the option “frown” ahead of a PDUFA announcement on a new drug, and a previously unreported informed jump model applicable to earnings announcements.