We use subhalo abundance matching (SHAM) to model the stellar mass function (SMF) and clustering of the Baryon Oscillation Spectroscopic Survey (BOSS) CMASS sample at $zsim0.5$. We introduce a novel method which accounts for the stellar mass incompleteness of CMASS as a function of redshift, and produce CMASS mock catalogs which include selection effects, reproduce the overall SMF, the projected two-point correlation function $w_{rm p}$, the CMASS $dn/dz$, and are made publicly available. We study the effects of assembly bias above collapse mass in the context of age matching and show that these effects are markedly different compared to the ones explored by Hearin et al. (2013) at lower stellar masses. We construct two models, one in which galaxy color is stochastic (AbM model) as well as a model which contains assembly bias effects (AgM model). By confronting the redshift dependent clustering of CMASS with the predictions from our model, we argue that that galaxy colors are not a stochastic process in high-mass halos. Our results suggest that the colors of galaxies in high-mass halos are determined by other halo properties besides halo peak velocity and that assembly bias effects play an important role in determining the clustering properties of this sample.