| Summary: | Limitations imposed by the traditional practice in financial institutions of running price and risk analysis on the desktop drive analysts to use simplified models in order to obtain acceptable response times. Typically these models make assumptions about the distribution of market events like defaults. One popular model is Gaussian Copula which assumes events are independent and form a "normal" (Gaussian) distribution. This model provides good risk estimates in many situations but unfortunately it systematically underestimates risk for unusual market conditions, the very time when analysts most need good estimates of risk. They run away from using Monte Carlo simulations since they can take days. We propose a Monte Carlo Simulation as a Service (MCSaaS) which takes the benefits from two sides: The accuracy and reliability of typical Monte Carlo simulations and the fast performance of running and completing the service in the Cloud. In the use of MCSaaS, we propose to remove outliers to enhance the improvement in accuracy. In the process of doing so, we propose three hypotheses. We describe our rationale and steps involved to validate them. We set up three major experiments. We confirm that firstly, MCSaaS with outlier removal can reduce percentage of errors to 0.1%. Secondly, MCSaaS with outlier removal is expected to have slower performance than the one without removal but is kept within 1 second difference. Thirdly, MCSaaS in the Cloud has a significant performance improvement over the Gaussian Copula on Desktop. We describe the architecture of deployment, together with examples and results from a proof of concept implementation which shows our approach is able to match response rates of desktop systems without making simplifying assumptions and the associated potential threat to the accuracy of the results.
|
|---|
