The traversal of transitions for regular execution of process models is controlled by the evaluation of transition conditions against possibly process data. To evaluate transition conditions against data during simulation requires the generation of data with appropriate assumptions about their value distribution. However, to specify these assumptions especially for complex and structured data is very challenging for the modeler.
Hence, the Simulation Feature allows you to simply provide probabilities for the transition condition becoming true. It is assumed, that such probabilities are reasonably easy to obtain based on some experience with real world data ("What is the percentage of credit applications having a total amount greater than $100,000").
Note, that even if you decide to use the data generation facility of the Simulation Feature, the generated data are not correlated with the transition traversal probabilities.
You can specify traversal probabilities in the Simulation > Probability property page in the property dialog for outgoing activity gateways:
Since it is possible to create a theoretically unlimited number of tokens from one process definition (e.g. using AND-Splits), this should be avoided during simulation. Limiting the number of loop iterations is possible using the Maximum Transition Traversals Per Process Instance configuration. It allows to specify the maximum number of traversals one process instance can perform for a certain transition. The default setting is Poisson 10.
For AND split gateways you may specify arbitrary percentages between 0 and 100% for each transition.
For XOR split gateways the sum the traversal probabilities for all transitions must be equal to 100.