Determining the scaling relations between galaxy cluster observables requires large samples of uniformly observed clusters. We measure the mean X-ray luminosity--optical richness (L_X--N_200) relation for an approximately volume-limited sample of more than 17,000 optically-selected clusters from the maxBCG catalog spanning the redshift range 0.1<z<0.3. By stacking the X-ray emission from many clusters using ROSAT All-Sky Survey data, we are able to measure mean X-ray luminosities to ~10% (including systematic errors) for clusters in nine independent optical richness bins. In addition, we are able to crudely measure individual X-ray emission from ~800 of the richest clusters. Assuming a log-normal form for the scatter in the L_X--N_200 relation, we measure sigma_ln{L}=0.86+/-0.03 at fixed N_200. This scatter is large enough to significantly bias the mean stacked relation. The corrected median relation can be parameterized by L_X = (e^alpha)(N_200/40)^beta 10^42 h^-2 ergs/s, where alpha = 3.57+/-0.08 and beta = 1.82+/-0.05. We find that X-ray selected clusters are significantly brighter than optically-selected clusters at a given optical richness. This selection bias explains the apparently X-ray underluminous nature of optically-selected cluster catalogs.