The other exception is stochastic optimization – for example, finding the right mix of stocks to maximise the 10th percentile of a portfolio’s return. That can take quite a while, and the software needs to have the memory to handle such large simulation data. For example, in capital allocation models for banks, a regulator can require such a high (illusory, in my view) precision of the 99.9th percentile, one has to run 500,000 samples or more. The first is when you need a high level of precision on a tail probability. There are two scenarios in which speed is important. When I started in this profession, a simulation of 500 samples would take a whole night to run but now all the products complete a run of 1-2,000 samples of a similar model in a couple of minutes. Simulation speed is generally not that important these days, and more a reflection of the quality of the underlying code. This topic describes the Monte Carlo simulation process in more detail. Excel is a great environment for running such models because Excel is so familiar and easy to build models with. That will tell us how confident we are in the results. The Monte Carlo simulation add-in allows the user to flag which cells are outputs, and then automatically generate several thousand scenarios by “running” the model so one can see the range and likely values of each output. ![]() The results are single-point estimates, and an obvious worry is – how confident are we in those estimates?Ī risk analysis add-in allows us to replace any of the spreadsheet’s input numbers with a function that generates random values that reflect how uncertain one is about that number – a process known as Monte Carlo simulation. These estimates can be built up from a set of calculations, perhaps over several sheets. Spreadsheets are often used to get an estimate of some future value, like the total cost of a project, the NPV of an investment or a sales forecast. How risk analysis models are built in Excel The main reason for using it is if you have no other choice, or you already have a perpetual license. In my view, is hugely over-priced compared to the other but gets away with it because it was the first to market and has locked in its users. That will double the cost, but the data mining capabilities are impressive. Also, though not reviewed here, Frontline Solvers offers a Data Mining tool that will complement the Analytics Solver optimization (which usually requires a lot of data to be meaningful). Analytics Solver is the best choice if you need to do stochastic optimization. I wouldn’t buy it.Īssuming that you have a budget, your choice is between ModelRisk, Analytic Solver and ModelRisk is the best risk modelling add-in for Excel by far, and the least expensive. I found it very hard to use and it is extremely limited compared with the other professional tools. Oracle Crystal Ball is a very old product and I doubt whether it will be much improved. If you have an extremely limited budget, RiskAMP is your only option but it is not of a professional quality.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |