Both (sorry!), there are known factors that the collisions are different, but also actual are actual statistical probabilities of different things that can happen when exactly the same initial condition occurs. And then there are even more random things like angles, the created particles can essentially fly in any direction after a proton-proton collision. Producing all those random options is one of the reasons we use very advanced simulations to predict what we expect. A substantial fraction of the grid computing power is used to produce those simulations.