I'm using the G2++ model, trying to fit the current term structure of interes rate. Accirding tio this model the choice of the enables to fit perfectly the current term structure. (see Brigo and Mercurio for example on chapter 4).What kind of errors should I expect from the model implementation?for instance I denote by error: where P(0,T) denotes the ZCB price for maturity T.According to our implementation we observe increasing percentage errors: up to the 3-5 years maturity we have percentage errors of approximately 100 bps, for longer maturity we have hugher errors, for instance the 30 years maturity has an error of approximately 10%...There could be some error in the model implementation or is that acceptable?Regards

It is an error in the model implementation.When you use enough simulation (say > 10.000), then you should get very small errors (usually < 0,1%).

Yes, many thanks. I found an error in the model implementation.What is according to you the right number of simualtion required?Keep in mind mind that we also employed a variance reduction technique (antithetic variate), and in this way we have the double number of simulation.Regards

Depends on which asset you are valuing with it. If you find the standard error of the option price low enough, than you have enough simulations.