We present a new measurement of the scaling relation between X-ray luminosity and total mass for 17 000 galaxy clusters in the maxBCG cluster sample. Stacking subsamples within fixed ranges of optical richness, N200, we measure the mean 0.1–2.4 keV X-ray luminosity, 〈LX〉 , from the ROSAT All-Sky Survey. The mean mass, 〈M200〉, is measured from weak gravitational lensing of SDSS background galaxies. For 9 ⩽ N200 < 200, the data are well fitted by a power law, 〈LX〉/1042 h−2 erg s−1=12.6+1.4−1.3 (stat) ± 1.6 (sys)(〈M200〉/1014 h−1 M⊙)1.65±0.13. The slope agrees to within 10 per cent with previous estimates based on X-ray selected catalogues, implying that the covariance in LX and N200 at a fixed halo mass is not large. The luminosity intercept is 30 per cent, or 2σ, lower than that determined from the X-ray flux-limited sample of Reiprich & Böhringer, assuming hydrostatic equilibrium. This slight difference could arise from a combination of Malmquist bias and/or systematic error in hydrostatic mass estimates, both of which are expected. The intercept agrees with that derived by Stanek et al. using a model for the statistical correspondence between clusters and haloes in a WMAP3 cosmology with power spectrum normalization s8= 0.85. Similar exercises applied to future data sets will allow constraints on the covariance among optical and hot gas properties of clusters at a fixed mass.