It’s surprising how often parts of models can be too simplistic. I’m the last person to advocate more detail in a model than is needed, but there is no point in a model that simply gives you back what you put in.How would you model the World Cup? The simplest model would be to sample from a distribution where each team is given a likelihood of winning – e.g. Brazil 28%. Running the model 1,000,000 times will show to your astonishment that Brazil are sampled the winner approximately 28% of the time.The model has told you nothing. This is an extreme case but be on your guard for similar inputs into your own models. In my experience, this happens more in ‘high level’ models where underlying causes of variability are not modelled in detail.