Sitting in an office on a sunny Saturday afternoon, this month's notes are focused on interesting events and some longer reads that I enjoyed. Probably I should be outside training for the Geneva Triathlon...
Upcoming Events in London
27.6.2019: Industrial Strength Data Science: A hell of a lot of skills
9.7.2019: Jason Hsu - Factor investing: Is it different in Asia?
15.7.2019: Asset & Liability Models that are useful in practice
5.-14.7.2019: British Summer Time Festival (Hyde Park)
IOSCO on regulatory considerations wrt. crypto-currency trading platforms
As regulatory agencies start to develop a taste for regulating cryptocurrencies, the International Organisation of Securities Commissions published a reasonably readable report on "Issues, Risks and Regulatory Considerations Relating to Crypto-Currency Trading Platforms". It's worth having a look at it to see the roadmap of future regulation.
The report can be found here: https://www.iosco.org/library/pubdocs/pdf/IOSCOPD627.pdf
Ben Hunt provides rather entertaining, while insightful reads on Epsilon Theory (https://www.epsilontheory.com/) - and thanks to Max for pointing out its existence. It's always a pleasure reading a refreshing perspective on current topics such as "the myth of college".
Estimating betas (i.e. exposure to risk factors) is hard. While empirical shrinkage factors (such as Vasicek 1971) have been introduced to adjust for over-/underestimated betas due to noise, Dr Lisa Goldberg did some very interesting work on formally deriving a more accurate (while still applicable) adjustment method. A useful read especially for those among you working in the asset management industry.
Here is the paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3282465
Computers are able to answer questions by now - it is fascinating to check see the progress shown on the Stanford Question and Answer dataset. The dataset consists of questions posed on a set of wikipedia articles, including a set of training answers. On the website, they also show the (surprisingly good) predicted answers resulting from the ML model trained on it.
Here is the set of questions and answers surrounding the Huguenots: https://rajpurkar.github.io/SQuAD-explorer/explore/v2.0/dev/Huguenot.html?model=nlnet (single model) (Microsoft Research Asia)&version=v2.0