Headspace

Observations:
1. I think it is impossible not to bring a voice recorder to a seminar. I can't write fast enough because there are just too many points to take down. I gave up and only jotted down interesting key points. I may have missed quite a lot
2. Going to conferences alone helps in networking. Most participants of the seminar came 'alone' they are very open to meeting people.

I'm really excited about two new things I've learnt. The first is loss data plotted in a double log graph in a talk by a BOJ Deputy Director General, Financial Systems and Bank Examination, Tsuyoshi Oyama. Paper was written by Tsuyoshi Nagafuki, and presentation available here as part of the discussion on the use of external data.

The second is a commercial use of EVT by a company called BPS. Instead of Gumbel copula, which according to the Very Important Dr, they use t-copula which results in a lesser capital amount. They work out the correlation of different tails for the individual business units to arrive at a more accurate (read: lower) capital computation. In a discussion with a colleague, he guessed that the correlations are risk loss correlations, while is natural to assume so since it works out similar causes of op risk, it is not intuitive how there can be different op risk correlations (high, medium, low) in one company when there is centralisation of operations and the presence of middle office teams. If so, the high low and medium risk reflects the skills of the operations and middle office team or even if the business unit is using the same team.