In this paper we are examining an error-detecting code which we have defined in our previous work. The code is defined using the algebraic structure quasigroup. If a quasigroup \((Q,*)\) is used for coding, then the code each information part \(a_0a_1\dots a_{n-1}\) (where for every i, \(a_i\in Q\) ), encode into \(a_0a_1\dots a_{n-1}d_0d_1\dots d_{n-1}\) where \(d_i=a_i*a_{i+1\; (mod \; n)}\) , for every i. Previously we have obtained the probability of undetected errors for all suitable for coding quasigroups of order 4 and identified the maximal number of faulty bits up to which the code surely detects the errors when for coding are used quasigroups from the top two classes in terms of the probability of undetected errors. Now, we will theoretically obtain this parameter for all other suitable quasigroups of order 4 and will conclude whether the quasigroups of order 4 that belong to the top classes in terms of the probability of undetected error also belong to the top class from the aspect of the maximal number of faulty bits up to which the code always detects the errors in transmission. With this we will obtain the quasigroups for which the code achieves optimal error-detecting capability.