A procedure is presented for estimating the probability of localized detection of an acoustic source under conditions of low signal-to-noise ratio (SNR). This technique, motivated by a tracking application, considers the strongest peaks in the ambiguity surfaces generated by matched-field processing. In an exhaustive series of simulations, conditions were defined under which a Gaussian cumulative distribution function (cdf) predicted the probabilities that a source at a particular location and SNR produced one of the m strongest peaks in the ambiguity surface, for various ranks m. The procedure for efficient performance estimation involved the use of noise-only data to define statistical thresholds for peaks in the ambiguity surface which corresponded to the various ranks. These thresholds could be used to predict analytically the probabilities of localized detection as a function of SNR. Results obtained for slanted and horizontal arrays indicate that this approach, combined with other analysis, can provide an effective way to estimate the probability that the source track is examined and to determine the detection performance and localization accuracy of arrays.