Which machine learning methods work best?

I am relatively new to machine learning, and it seems that given a new problem to tackle, e.g. those on Kaggle, the best approach is to try several of the existing methods and see which one works best. Decision trees are great because they are simple, less likely to go bad, and easy to visualize. They also often give the best result. Is there a way to know up front which method would be best for a given problem?