Subj. This seems to me to be a wrong approach, moreover this drives to such mistakes:

AdaBoostClassifier(algorithm="SAMME", base_estimator=SVC()).fit(trainX, trainY).predict_proba(testX)
---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
 in ()
----> 1 AdaBoostClassifier(algorithm="SAMME", base_estimator=SVC()).fit(trainX, trainY).predict_proba(testX)

/Library/Python/2.7/site-packages/sklearn/ensemble/weight_boosting.pyc in predict_proba(self, X)
    716             proba = sum(estimator.predict_proba(X) * w
    717                         for estimator, w in zip(self.estimators_,
--> 718                                                 self.estimator_weights_))
    719 
    720         proba /= self.estimator_weights_.sum()

/Library/Python/2.7/site-packages/sklearn/ensemble/weight_boosting.pyc in ((estimator, w))
    715         else:   # self.algorithm == "SAMME"
    716             proba = sum(estimator.predict_proba(X) * w
--> 717                         for estimator, w in zip(self.estimators_,
    718                                                 self.estimator_weights_))
    719 

/Library/Python/2.7/site-packages/sklearn/svm/base.pyc in predict_proba(self, X)
    493         if not self.probability:
    494             raise NotImplementedError(
--> 495                 "probability estimates must be enabled to use this method")
    496 
    497         if self._impl not in ('c_svc', 'nu_svc'):

NotImplementedError: probability estimates must be enabled to use this method