The right balance of forward rays, backward rays, and refinement passes will depend on simulation accuracy needs and your PC. As with the other Monte Carlo simulation types, increasing the number of rays will improve accuracy, but will also increase simulation time and memory usage. Another consideration is the ratio of the number of forward rays traced and the number of backward rays traced. As a rule of thumb, the number of forward pass rays should be larger than the number of backward pass rays traced.
A unique control of the hybrid simulation method is a refinement pass. A refinement pass is the number of forward/backward iterations to perform, and can be used to manage the amount of memory used during simulation. Each refinement pass will allow you to effectively increase the number of forward pass rays traced without increasing the memory requirements. If you need additional accuracy but your PC is limited on memory, try increasing the number of refinement passes, while leaving the forward pass rays and backward pass ray settings the same.