The success of a learning algorithm depends on its ability to balance between fitting the data and avoiding overfitting.
In business, you have to be prepared to take risks, but calculated ones.