Incorrect :) What you get using DQN is exactly the probabilities to select each action. These probabilities are based on the predicted Q-Values.