The main aim of this paper is to devise a reliable, cost effective Fault Tolerant Central Valuation Scheme. A qualitative reasoning for the high failure rate of the candidates is represented by a fault tree. A genetic algorithm is used to model the evaluation system in such a way that optimum number of examiners are oppointed to improve the overall reliability of the valuation scheme subject to the cost constraint. A systematic representation of the problem is given in steps such that genetic algorithms can be applied. This paper also suggests the optimum load for the examiners (number of papers to be corrected per day) and points out the number of redundant examiners per subject to be appointed to maintain reliability at the highest level as possible.