Xgboost predict_proba

Hi all,

I training my xboost model and used predict_proba to get probabilities for my binary classification problem, but the output score is between 0.19 and 0.76, so I wonder what does it mean that I did not get score from (0,1) but narrower range?

Here is my code, which is rather super standard:

from xgboost import XGBClassifier

clf = XGBClassifier(seed=42)

max_depth = range(3, 15, 1)
min_child_weight = range(5, 12, 1)
subsample = [i / 10.0 for i in range(6, 10)]
colsample_bytree = [i / 10.0 for i in range(6, 10)]
learning_rate = [0.0001, 0.01, 0.1, 0.2]

# Create the random grid
random_grid = {'max_depth': max_depth, 'min_child_weight': min_child_weight,
               'subsample': subsample, 'colsample_bytree': colsample_bytree, 'learning_rate': learning_rate}

clf = random_grid_search_for_clasiffier(clf, random_grid, scoring_function, 
                                        X_train, y_train, X_test, y_test, n_iter=1500)

y_results = clf.predict_proba(X_test)


why would you think you would get props in 0…1 range? 0 would happen only if given value was never seen in training and 1 in the opposite case.

Sure. I was maybe not 100% precise but I mean that they should be distributed between 0 and 1 and i my case I have props more less between 0.2 and 0.8