If the optimal number of trees for a model is 500, but we accidentally trained both a random forest and boosting with 1000 trees, which model is more likely to overfit and why?