Latest Practical CVA & Counterparty Credit Risk Advancements
09.00 - 10.30: Marco Bianchetti: Head of Financial Modelling & Validation, Intesa Sanpaolo & Stefano Scoleri: IBM Italia
Computing Risk Measures: Quasi Monte Carlo vs Pseudo Monte Carlo 10-1
- Quasi Monte Carlo in finance: state of the art
- Why QMC is so good ? Effective dimension and global sensitivity analysis
- Computation of prices, greeks, and risk measures
- Performance analysis: convergence, error, stability.
Break: 10.30 - 10.50
10.50 - 11.40: Jörg Lotze: Technical Lead & Co-Founder, Xcelerit
CVA Implementation Secrets: Lessons from the Trenches
This talk will give a practitioner's view of developing and deploying CVA and CCR algorithms to run at high speed on GPUs and other acceleration hardware.
- CVA and CCR Algorithms: Where is the complexity?
- Design Choices - Common Mistakes
- Where are the software performance bottlenecks?
- Introducing GPUs and other accelerators: What do they do well?
11.40 - 12.30: Dimitris Karyampas: Director IB Exposure Measurement, UBS
Backtesting Counterparty Credit Risk Models
- Historical vs Market Implied approach
- Test-Statistic properties
- Capital Buffer
- Real Examples
Lunch: 12.30 - 13.30
13.30 - 15.00: Norbert Hari: Head of Quantitative Development Team, ING Bank
Challenges in Counterparty Exposure Modeling
Counterparty Exposure Management System running on GPUs
GPU Computing – What and Why?
- Wrong-way / right-way risk
- Additional Termination Events
- Benefits of combining Numerical Methods: MC, PDE, LS, SBGM, COS etc.
- Efficient sensitivity calculation
- Impact of single- vs multi-factor models
- Impact of smile
Break: 15.00 - 15.15
15.15 - 16.45: Luis Manuel García Muñoz: Head of Interest Rates, Credit and CVA Quant Teams, BBVA
CVA for Credit Derivatives
We will deal with the calculation of CVA of portfolios of credit derivatives, putting special focus in the way in which default times of the different credit references are correlated.
- Weaknesses of traditional copula approaches.
- The Marshall-Olkin copula: a feasible approach for a reduced number of credit references.
- Nested Archimedean copulae: a way to reduce the number of parameters for multidimensional Marshall-Olkin copulae.
- Relationship with the Lévy-frailty model.
- Correlation skew produced by canonical Lévy subordinators (constant drift plus Poisson process).
- Correlation skew produced by a constant drift plus a compound Poisson process with exponential jumps.
- Impact of stochastic default intensities in the correlation skew implied by the model.
- Calibration of the model to credit index tranches.
- Impact of the different parameters on the CVA of a portfolio of CDSs.
End of Conference
Receive an extra 5% discount when booking 3 or more delegates. Plus Register to the Main Conference + Workshop and receive a £150 discount. To register click here!