Is there a mistake in the Python Package Introduction?

In the Setting Parameters section of the Python Package Introduction of XGBoost documentation, the code for specifing validations set to watch performance goes like this:

evallist = [(dtest, 'eval'), (dtrain, 'train')]

I followed this code and found early stopping rounds did not work, as it was targeting the train score rather than the eval score. It seems that the order should be reversed:

evallist = [(dtrain, 'train'), (dtest, 'eval')]