ECOC scores better accuracy in binary classification than SVM

Hi,

I have a binary text classification problem.

I used an ECOC Model using linear svm and I used just an svm with a linear kernel (default). Even though it is a binary problem and I usually wouldn't use an ECOC for that, the ECOC model consistently outperforms the SVM by a small margin.

I use the following code:

svm = fitcsvm(Train,ID,"Kfold",5);
mdl = fitcecoc(Train,ID,"Kfold",5,"Learners","Linear");

Shouldn't they both yield the same result on average?

Thank you very much in advance!

Best regards

Malte