We calibrated the crop model STICS and used it in a simulation experiment to assess the effect of no-tillage cropping systems, with and without millet as cover crop (NTM and NT), on the water-limited yield of maize, its inter-annual variability, water runoff and drainage below the root zone, compared to conventional tillage-based cropping systems (CT). We ran simulations for two contrasting climates in the Brazilian Cerrado, Goiânia with 1600¿mm of annual rain and Barreiras with 1160¿mm, using a series of historical weather data and considering two soil types with different soil water storage capacities (i.e. different maximum rooting depths of 90 and 180¿cm). We identified the best suitable sowing windows based on agronomic (yield and yield variability) and environmental (runoff and drainage) performances of the cropping systems. Compared to CT, NT and NTM systems increased water-limited maize grain yield and its stability. For example, for the simulations with 90¿cm set as maximum rooting depth, grain yields for CT, NT and NTM, averaged over all sowing dates, were 3.41, 4.35 and 4.28¿t ha-1 in Goiâna, and 7.02, 7.67 and 7.5¿t ha-1 in Barreiras. No-tillage systems markedly decreased water runoff and hence erosion risks. For example, simulated runoff was on average 4% of the annual rain for NT and NTM against 13% for CT at both locations. In contrast, no-tillage systems increased water drainage below the rooting zone, hence with risks of increased nutrient leaching. For example, for the 90¿cm maximum rooting depth, simulated drainage was 48% and 41% of the annual rain for NT and NTM against 32% for CT in Goiânia, and 29% and 26% against 17% in Barreiras. Reconciling the objective of maximizing yield and its stability with that of minimizing both water runoff and drainage was only achieved under no-tillage with millet as cover crop (NTM) for a very narrow range of sowing dates. Sowing all fields of the farm within a narrow optimal window requires, however, important