The study aimed at comparing impact of difference in item selection procedures on ability estimation precision and Relative Efficiency of achievement test. An achievement test composed of (60) multiple choices items set within Educational Evaluation course and applied to (409) student at faculty of Education. Item calibration was done using Three-Parameters Logistic Model. Three tests were derived from the overall test upon theta at maximum information (θmax), maximum information (I(θmax) and classical discrimination (Dclss). Each one composed of (20) items that were calibrated by (3PLM) model. Results: The three tests are equivalent in psychometric properties, statistical indicators and (θ estimates) with the overall test. No differences in ability estimation between the overall test and three tests. There are differences in the accuracy of ability estimation for the second test (I(θmax) and third (Dclss) comparing first (θmax). No differences in accuracy of ability estimation between the second test (I(θmax) and third (Dclss). Tests were arranged upon values of (RE) as of the highest accuracy as follows: the second (I(θmax), third (Dclss) and lastly the first (θmax) with wide convergence between the indicators of the second (I(θmax) and third (Dclss).