Computer simulation is the only realistic method of evaluating alternative methods of breeding hybrid forest trees. Empirical tests would be very long term and expensive. This paper describes the development of a simulation program, called XSIM, which generates two different but closely related outcrossing tree species. The genetic correlation between performance in each parental species and performance in the resulting hybrid can be set, in addition to the amounts and types of variances in each parental species. The breeding strategies available for testing include conventional reciprocal recurrent selection, reciprocal recurrent selection with forward selection, recurrent selection within each pure species, and the creation of a synthetic species. XSIM allows the strategies to be compared using the same base populations, equivalent selection intensities, and comparable mating patterns. Innovative best linear unbiased prediction procedures allow all ancestral and current progeny generation data, from both parental species and the hybrid, to be analysed together. The theoretical basis for the simulation is given, and genetic and statistical models are described. In summary, XSIM allows rigorous comparisons of the strategies in terms of genetic gain per time and provides useful insight into hybrid forest tree breeding.