To be fair. Even using data can be a problem. For instance, if there is no standard test data available for a plane, one must resort to the use of other sources to fill in the blanks so a proper flight model can be derived.
The problem with is the "cherry picking" which can occur. Take 20 books about one plane and many people will take the best parameter from each book and claim that is what should be used to model flight for a plane.
Of course, when you run the numbers, they do not come out right due to the data conflicting with each other as each book usually has done a decent job with making sure their numbers do compliment each other.
Without manufacturing test data, you can only hope to pick the data set you hope is correct. Fortunately, there is pretty good data for most planes that flew in World War II. None of that data is pilot anecdote, by the way.