Link adaptation in which the transmission data rate is dynamically adjusted according to channel variation is often used to deal with time-varying nature of wireless channel. When channel state information at the transmitter (CSIT) is delayed by more than channel coherence time due to feedback delay, however, the effect of link adaptation can possibly be taken away if this delay is not taken into account. One way to deal with such delay is to predict current channel quality given available observation, but this would inevitably result in prediction error. In this paper, an algorithm with different view point is proposed. By using conditional cdf of current channel given observation, outage probability can be computed for each value of transmission rate $R$. By assuming that the transmission block error rate (BLER) is dominated by outage probability, the expected throughput can also be computed, and $R$ can be determined to maximize it. The proposed scheme is designed to be optimal if channel has ergodicity, and it is shown to considerably outperform conventional schemes in certain Rayleigh fading channel model.