Petrophysical properties of tight sandstone reservoirs are complex which brings difficulties to fluid discrimination. Rock physics makes it possible to obtain petrophysical properties from elastic parameters. However, both deterministic rock physics and statistical rock physics have corresponding limitations. By combining deterministic rock physics and statistical rock physics, a joint posterior probability is proposed for fluid discrimination. To consider the effect of complex pore structure and permeability in tight sandstone reservoirs, a new deterministic rock physics model is built. In this model, soft porosity and connected porosity are quite important parameters to describe the above-mentioned reservoir characteristics. Assuming the noise follows a Gaussian distribution, we can obtain the posterior probability of gas saturation from the deterministic rock physics. Bayes discriminant is an effective method for statistical rock physics to estimate the prior, condition and posterior probabilities of petrophysical properties from well-logging data. Thus, the posterior probability of gas saturation belonging to the statistical rock physics is obtained. To guarantee the accuracy of fluid discrimination, the reflectivity method is used to achieve high-precision elastic parameters from seismic data. Application examples of well-logging data and seismic data confirm the validity of the proposed joint probabilistic fluid discrimination.