fitrsvm optimized with cross-validation

To whom it may concern,

sorry for bothering you with a super silly question, but I cannot understand a detail about the example "Optimize SVM Regression" shown in the documentation of the fitrsvm function. As in the regression setting, what does the column "BestSoFar (estimated) mean in the output of function fitrsvm? I understand that BestSoFar (observed) is the minimum of the objective function log(1+cross_validation_loss), but how is it "estimated"?

Sorry for bothering you and many thanks for your attention! Best,

Fabio.