In southern New South Wales, wheat is sown in the autumn (April-June), usually after a period of fallow to conserve soil water. In order to predict the effect of fallow and of new tillage systems, in particular no-fallow followed by direct drilling, on total available soil water (ASW) at sowing and on moisture-based sowing day probabilities, a simulation model combined with historical daily weather data was used. The model comprised a fallow subroutine with no plants, a no-fallow component involving weed germination, growth and water use, both followed by a wheat crop phase for continuous simulation. The model was calibrated and satisfactorily validated using data from several tillage experiments conducted in the region over the period 1979-1984. It was then used to compare long fallow (LF, 1 September start), short fallow (SF, post-wheat-maturity start, November-December), late short fallow (SFl, 15 February start) and grazed no-fallow (NoFg) over 41 years at Yass (684 mm rainfall), Wagga Wagga (566 mm) and Griffith (420 mm). On average, ASW on 1 May was 88 mm (LF), 61 mm (SF), 52 mm (SFl) and 33 mm (NoFg), with greater fallowing gains the wetter the site. Sowing probability in May averaged 77%/day for all fallow treatments and 54%/day for grazed no-fallow. In a number of years, especially following no-fallow, there were few sowing days in May. The average green herbage (weeds) consumption by sheep (at 10 animals/ha) on NoFg from harvest to 1 May was only 33 g/m2. Results were insensitive to the assumed starting weed leaf area index (range 0·001-0·025) after weed-germinating rain, and to stocking rate (range 0-40 per hectare), but were moderately sensitive to the upper and lower limits of seed-bed water content for sowing. Model assumptions are discussed and it was concluded that quantification of the likely cost of not fallowing in terms of reduced ASW and reduced number of sowing days was possible. © 1990.