Since the total evidence you can get from examining all previous guesses (assuming conventional strategy and rewards as before) gives you only a 4/5 accuracy, and you can get 2/3 by ignoring all previous guesses and looking only at your own draw: Yes, rewarding correct contrarians at least 20% more than correct majoritarians would provide enough incentive to break the information cascade. Only until you've accumulated enough extra information to make the majoritarian answer confident enough to overcome the difference between rewards, of course, but it would still equilibrate at a higher accuracy.