Related Topics
Risksvr™ gives you the modeling power to optimize returns while minimizing
exposures against all the different risk factors that can affect your bottom
line and assess how much Market, Credit, Liquidity and even modeling
risk you can sustain before these events actually impact your bottom line either as a
balance, a collateral value, a series of limits, a percentage or even a rating !
In today's volatile marketplace RiskServers' Financial Risk Management
Calculation Server can easily make the difference between consistent returns and
unsustainable losses!
Market Risk
Risksvr™ covers the four Methodologies standard to Risk industry:
These four methodologies rely on the evolution of all related market risk factors: Individual Interest Rates Vertices within and across base Term Structures, Spread curves, Foreign Exchange Rates, Equities, Commodities, Forward Rates Volatilities and Option Implied Volatilities Surfaces.
Risksvr™ aggregates results directly in complex container. This means you can associate any computed value with one or multiple tags and then proceed to "transform" by slicing and dicing absolute and marginal or incremental results into reports that are deemed strategic to your business.
This approach offers limitless possibilities in terms of risk measurements reports.
Liquidity Risk To measure Liquidity risks, Risksvr™
incorporates price slippage either as a number of days until the trade will be
unwounded or a bid-ask spread that evolves stochastically throughout time: Credit Risk To measure Credit risks, Risksvr™ applies
a random counterparty default model based on stochastic recovery rates, default
probabilities and generalized expected default frequencies. Credit Curves Expected default probabilities Marginal default probabilities or
Hazard rates. Survival Probabilities. Transition matrices that evolve
throughout time. If one single matrix is provided then Risksvr will assume
it will be used to build the entire curve Transition and Migration Correlated and
uncorrelated default methodology: Direct Default
Non-Default Observations
Correlated: As with
standard obligor correlation, Risksvr expects a weighting scheme for each
risk factor associated to a counterparty. Time-to-Default Copula based methodology: Risksvr™
offers the mechanisms to measure bilateral and multi-lateral netting across
account hierarchies. Every counterparty can hold nettable and non-nettable
accounts with their own specific collateral and currency. Country Growth and Contraction Risksvr™ can simulate
country risk devaluation, contraction or growth (Sovereign political risk, Inflationary pressure, Devaluation
policies, etc.) by computing both Country Exposures and Country Losses. Cash Accounts, Overnight Funding and Collateral. Risksvr™ can measure
funding gaps and long term profitability scenarios of positions. You can therefore take into account movements of collateral posted to fund these
positions as well as estimate the impact of reinvestment policies and this
either from a non-guaranteed or guaranteed counterparty relationship Generic Reporting, Modularity and Transparency Risksvr™
is designed
as a generic modular engine. Better yet, every process can be
overridden by user defined input data, This means you have complete control over
each calculation step in the engine. Each step can be captured and replayed, which greatly facilitates
replication, validation and benchmarking. For example, you can decide to override
the internal engine and inject your own Random data, your own forward rates,
your own Zero Coupon Rates, your own Factored Matrix, your own unitized
returns, your own Pricing Routines, your own Marginal / Conditional
Default Probabilities, your own Counterparty factorized weights and mappings,
etc. You can also switch between unlimited sources of market data (file,
sockets, db) (see Template Data / Benchmarking) which greatly
facilitates checking of results. Associative Tags Risksvr™
accumulates results according to tags and dimensions In Risksvr™ parlance, a tag is a
token in a URL like
hierarchy associated to one or a group of trades. Tags can be sliced and diced as well as "custom operated". Embedded Combined Marginal and Incremental
Measures By design Risksvr™
accumulates marginal sensitivities of the trade's contribution with regards to
other holdings that form a position and portfolio on the fly. This means
incremental VaR or Losses as well as any marginal statistics are always
available for any report. This has important consequences in terms of
stress test "hot-spots" identification, hedge analysis, real-time value-at-risk,
default loss or credit enhancement and portfolio diversification benefits. Engine Granularity Risksvr™ comes pre-configured
with industry standard settings and assumptions. This makes the
engine very easy to use for beginners and yet amazingly sophisticated for
leading edge users or researchers that want to investigate new approaches.
Designed In advance for Cloud Computing and beyond Risksvr™ was
designed with the underling concept that has now become know as cloud computing. Risksvr™
doesn't need any special middleware to run from clouds as the
kernel was designed with this approach many years before the term
"cloud" was even coined ! This
means the same Risksvr™ exectuable
can be deployed as a web server or stand-alone calculation server
which can run from numerous local and or remote data sources and
data sinks.
Each piece of information supplied to the engine can
come from a different repository, with it's own data schema, layout
and format. This means you can very well source encrypted trade
information from a database located in a subsidary over a public
network, while other trades orginiate from an xml file located on a
VPN and others come from a spreadsheet running on your computer.
While Equity Market Data could come from a public repository located
on the internet and yield curve data from a proprietary web server
and foreign exchange fro a flat file, etc. The same holds true for output. The engine is
designed to generate multiple layout formats and types. This means
you could very well generate html or pdf reports on the fly or store
every result in multiple databases from different vendors which are
then used to validate results and measure excessions.
In this respect configuration opportunities are infinte and can
adapt to the most complex environment, by design!
This makes the engine very
easy to use for beginners and yet amazingly sophisticated for
leading edge users or researchers that want to investigate new
approaches.
Selected values can then be backtested automatically against mark-to-market
values in order to confirm quality of selected model and confidence
multiplier.
Risksvr uses default probabilities from a credit curve when simulating default
mode.
Expected default frequencies are computed from the marginal default
probabilities extracted from the credit curve.
The credit curve can be built from:
Risksvr™ extends beyond the fixed 1 year horizon and provides a model that
can incorporate the full term structure of default probabilities, default
correlations, hazard rates and / or transition matrices.
Risksvr can also use Transition Matrices to migrate counterparty
ratings over simulation horizons.
Migration opens the door to two types of migration Migration up-to-default and
Migration with Full cost of carry:
Risksvr™ incorporates tow different default simulation technologies:
Univariate / Independant Default: If no
weightings are provided, the engine falls back to the default mode, which
applies no correlation between the counterparty's probabilities of default. Instead Risksvr
computes the likely-hood of default from a random uniform draw. If default
is observed then recovery is simulated stochastically and then applied to
the agent's exposure in order to compute LGDLoss-Given-Default)
This weighting is usually assumed to follow a country index approach, but
any weighting scheme can apply. The
simulation is then weighted and mapped to the Normal distribution's cumulative density. In
its most simplistic form, this model falls back to the CreditMetric(tm) model.
In
it's fully fledged form, this allows for migration between simulation
horizons with different transition matrices. This model expected a rating
assigned to each spread curve in order
to compute the full impact of upgrades and downgrades as they take place
over the simulation horizons.
If this is not enough you can actually switch the defaults engines to uses
copula based Time-to-Default or Time-To-Survival methodologies instead of
the more direct edf approach.
This can apply for both Counterparty Risk (Counterparty Credit Exposure and
Country Based Exposure) and Obligor Based Risk (i.e. Credit Sensitive
Assets).
Full account netting
Each position can associate cash-accounts to individual exposures (i.e. legs).
Each exposure can then credit/debit the cash-account according to
the proceeds or costs associated with the positions (coupons, dividends (if paid
in cash), payments, option exercise, etc.
Cash-accounts can thus mimic very closely what happens in practice.
Each component is connected to the other via generic streams.
Reports can thus be produced for every piece of information processed by
Risksvr(Tm).
Tags support the same syntax as conventional URLs. (e.g.: /portfolio/fx/usd
could be a valid tag). Tags are always associated to a dimension. If no tag is
defined, the engine assumes the tags belong to the same default dimension.
Tags come in many flavors and types (market, credit, factor, time-band,
benchmark, limits, hedge-accounting, etc)
However, every underlying assumption that drives each
Methodology can be changed by the user or administrator, if need be.
This approach is based on the idea that if you don't ask, then you probably don't need
to know....
This obviously makes your life a lot easier, since you don't need to deal with
complex setup values which only Risk Management professionals are accustomed to.
On the other hand you can always change these values if you want to tweak results
according to your own assumptions.
The engine is based upon generic financial concepts that can be relaxed or
constrained in order to create entire sets of assumptions that steer each
methodology.
©MMXII 2012© RiskServers SA. All rights reserved.
div>