Increasing the selectivity of sensors, while at the same time reducing their complexity, size and cost, are challenges to the sensing community. To this end, an area of exploration has been the development of filter-based chemical sensors. We have recently introduced an approach that utilizes multiple, broadband, infrared (IR) filters to enable discrimination of target chemicals, in the presence of potential interferents that have IR spectral signatures in a limited waveband. Our analysis technique, comparative discrimination spectral detection (CDSD), utilizes a set of broad IR transmission filters, to discriminate between a specific target chemical and multiple interferents with strongly overlapping IR spectra. We have demonstrated the ability of this technique to correctly distinguish between chemicals in the carbon – hydrogen stretch region of the IR absorption spectrum (2700 – 3300 cm-1; 3.0 – 3.7 μm). We present a numerical study exploring the choices of desired optical filter sets, and the resulting overall discrimination by these filter sets. Filter parameter choices, such as the peak transmission position and bandwidth, are fundamental in filter-based chemical sensing discrimination systems. In this paper, we describe a systematic numerical approach used to explore how optical filter properties, and filter overlap affect corresponding discrimination results. We describe the interaction between the overlapping spectra and various filter sets on both target and interferent chemicals. We discuss which filter parameters provide optimum selectivity for specific target chemicals and how this information can be utilized to select filters for future direct-filter sensors based on this methodology.