An index coding scheme in which the source (transmitter) transmits binary symbols over a wireless fading channel is considered. Index codes with the transmitter using minimum number of transmissions are known as optimal index codes. Different optimal index codes give different performances in terms of probability of error in a fading environment and this also varies from receiver to receiver. In this paper we deal with optimal index codes which minimizes the maximum probability of error among all the receivers. We identify a criterion for optimal index codes that minimizes the maximum probability of error among all the receivers. For a special class of index coding problems, we give an algorithm to identify optimal index codes which minimize the maximum error probability. We illustrate our techniques and claims with simulation results leading to conclude that a careful choice among the optimal index codes will give a considerable gain in fading channels.