Basis Curves

Basis curves can be modeled as a spread above or below another curve. For example the AA swap curve (the "target" curve) can be modeled as the AAA swap curve (the “underlying” curve) plus the AA spread (a small margin above the AAA curve).

To achieve this you need three curve definitions plus one basis curve definition. The basis curve definition binds the underlying and the spread curves together to create the target curve.

In the above example the basis curve definition basically says that:

MM.ZERO.AAA.USD + MM.SPREAD.AA.USD = MM.ZERO.AA.USD

Whenever a pricing model asks for a rate from the MM.ZERO.AA.USD curve the system will first look for actual rates, and having not found any will fall back to using the basis curve definition to generate the rate from the underlying and spread curves.

The mechanism used for retrieving a rate from a basis curve is that the system retrieves the underlying rate from the underlying curve (using interpolation if necessary on the underlying's term structure) and the spread from the spread curve (using interpolation if necessary on the spread term structure), and then adds the underlying to the spread to give the target rate.
 
So this means that when a pricing model asks for a rate from the target curve during a Monte Carlo simulation the system will retrieve the appropriate rate from the underlying curve and add the spread. By assigning a sensible stochastic process (eg lognormal) to the spread curve the underlying and target curves are evolved in a realistically correlated fashion on each simulation path.


Derived Curves

Certain curves can be derived from other curves, not by the simple application of a spread (as in basis curves described above) but via a mathematical model.
Two examples are the derivation of discrete dividends built from the zero curve, equity price curve and discrete dividend curve, and the creation of a survival curve from the zero curve, CDS spread and recovery rate.

Rate Source Details

The security database has a list of valid rate sources for an org. This is checked whenever a calculation tries to use rates or parameters.
The Assumptions configuration file contains a rate sources priority list consisting of blocks of wild carded curve keys for each rate source. For the organisation Ajax an example list might be:

Ajax
MM.ZERO.SWAP.USD

Xignite
MM.ZERO.SWAP.*
MM.ZEROVOL.CAP.*

Ajax
*.*.*.*

Reuters
*.*.*.*

Every lookup of a curve (on any date) or parameters or correlations for a curve will use the above list to find the highest priority data. So it is possible that a history set for a curve to be used for historic simulation (or parameter or correlation estimation) will be composed of rates from multiple sources. If this creates a problem for a customer they can load their own history set.
Whenever an org loads its own rates or calculates its own parameters or correlations these are stored with the org’s own rate source which would normally be set as highest priority for the org.