It depends on your priors. If you set it up properly the evidence for either model from one sample would be weak. You shouldn't be confident in either model after one trial, but you can eliminate some models, i.e. all balls are green.