Minimax Optimality In Online Learning Under Logarithmic Loss With Parametric Constant Experts