As long as the code I'm writing matches the training data it is always correct.

Then it's likely an opportunity for factoring. The ideal is to never have duplicate code