Open-Source Project for
Credit Risk Modeling
Frequently Asked Questions
- Why this project name ?
- What is this logo ?
- What is the project's maturity ?
- Is there a company that provides support for CCruncher ?
- What about performance ?
- Why can the input files be gzipped (*.gz) but not zipped (*.zip) ?
- Some samples exhibits strange behavior. What is happening ?
CCruncher software uses a numerically intensive procedure to quantify the portfolio credit risk. These kinds of programs are called 'number crunchers'. The mix between the terms 'credit' and 'number cruncher' generates the name 'credit cruncher'. CCruncher is not related to the recent credit crunch, but is a tool that can aid to control the risks involved in this crisis.
CCruncher's logo is based on a sculpture by Robert Llimós, a catalan artist. The original work is named Threshold and is located in Atlanta. It has a twin version named Marc that is located in Barcelona.
ccruncher-2.4.1 is a stable version. If you have any bug or suggestion to report, please send an email to CCruncher team.
Yes. Tatine.es provides consulting services for the analysis of SME loan portfolios using CCruncher. It also provides support services such as training, parameters estimation, integration, maintenance or ad-hoc development related to CCruncher.
CCruncher is designed to be extremely fast. When there are 10 obligors, it computes 1000000 simulations in few seconds. When obligors are 50000 it takes few minutes to do the same task. CCruncher execution time depends on multiple factors: machine (processor, L1/L2 caches, cores, etc.), number of simulations, number of obligors, number of ratings, number of factors, number of segmentations, and number of assets.
The zip algorithm is covered by patents, but gzip is not; furthermore gzip is open source. You can download a gzip tool for windows from 7-zip. UNIX environments have the gzip/gunzip commands out of the box. You can read more at the gzip website.
1) Standard Error of VaR increase when the number of simulations increases:
We use the Maritz-Jarrett method to obtain these values. To reduce the
computational cost we truncate the involved sum. In some cases, this action
causes the observed behavior.
2) VaR plot has jumps: This happens because the resulting portfolio loss distribution is discrete with a support composed of few values.