ABSTRACT The neutral hydrogen (H i) gas is an important barometer of recent star formation and metal enrichment activities in galaxies. I develop a novel statistical method for predicting the H i-to-stellar mass ratio, $f_{\mathrm{H\,{\small I}}}$, of galaxies from their stellar mass and optical colour, and apply it to a volume-limited galaxy sample jointly observed by the Sloan Digital Sky Survey and the Arecibo Legacy Fast ALFA survey. I eliminate the impact of the Malmquist bias against H i-deficient systems on the $f_{\mathrm{H\,{\small I}}}$ predictor by properly accounting for the H i detection probability of each galaxy in the analysis. The best-fitting $f_{\mathrm{H\,{\small I}}}$ predictor, with an estimated scatter of 0.272 dex, provides excellent description to the observed H i mass function. After defining an H i excess parameter as the deviation of the observed $f_{\mathrm{H\,{\small I}}}$ from the expected value, I confirm that there exists a strong secondary dependence of the mass–metallicity relation on H i excess. By further examining the 2D metallicity distribution on the specific star formation rate (sSFR) versus H i excess plane, I show that the metallicity dependence on H i is likely more fundamental than that on sSFR. In addition, I find that the environmental dependence of H i in the local Universe can be effectively described by the cross-correlation coefficient between H i excess and the red galaxy overdensity ρcc = − 0.18. This weak anticorrelation also successfully explains the observed dependence of H i clustering on $f_{\mathrm{H\,{\small I}}}$. My method provides a useful framework for learning H i gas evolution from the synergy between future H i and optical galaxy surveys.