Riskdata Monte Carlo VaR methodology
by Riskdata, 2015
VaR is an estimation of the maximum amount that a security, a portfolio of securities, or an index, may lose at a given time horizon for a given level of confidence. For example, the “1 day 99% VaR” of the S&P500 index being equa to 4% means that the estimated probability of the S&P falling more than 4% over the course of a day is less than 1%.
Riskdata proprietary HuLK VaR model has been developed to overcome the possible over- or underestimation of the risk during a temporary market crisis. As a matter of fact, it provides a more responsive estimate in that it reacts swiftly to a change of market regime. It also attempts to anticipate increases and decreases in VaR numbers by identifying micro-signals which can be often be seen in pre- or post-shock periods. Backtests performed on shorter time periods (one year or even a few months) show that the frequency of exceptions during turmoil periods is more in line with the specified VaR level than when considering traditional Monte Carlo VaR models.
The behavior of HuLK VaR in the various historical crisis and, in particular, through the Credit Crunch in 2008 has shown how effective it is at even anticipating crises, rather than simply reacting to them.
While these may all be good reasons indeed for why our proprietary VaR model bears its name, it is the very nature of the market and the underlying probability laws governing it which are at the root of this choice, as will be revealed when diving into our methodology.