Average daily gain (ADG) is an important indicator of cattle performance, but frequent liveweight measurement remains difficult under commercial grazing conditions. This study investigated whether accelerometer-derived behaviour and activity measures, combined with animal-level and meteorological variables, could be used to predict ADG in grazing cattle. Two month-long field experiments were conducted in Queensland, Australia, in 2022 and 2023, involving 120 Brahman and Droughtmaster steers fitted with triaxial-accelerometer collars. Animals were weighed approximately weekly, and interval-average daily weight gain was calculated between consecutive weighing occasions. Accelerometer data were used to derive daily activity and time budgets for grazing, ruminating, resting, walking, drinking, and other behaviours. Regression models were trained to predict liveweight gain using these measures, together with breed, initial liveweight, and meteorological variables, and were evaluated using grouped five-fold cross-validation at the animal level. Random forest achieved the best numerical performance, with a mean $R^2$ of 0.4814 and RMSE of 0.7398~kg/day, although its performance was nearly identical to that of linear regression ($R^2=0.4788$, RMSE$=0.7413$~kg/day). These results indicate that the combined accelerometer-derived, animal-level, and meteorological variables captured a moderate proportion of the variation in interval-average ADG. Integrating wearable-sensor data with additional nutritional, biological, and environmental measurements may further improve cattle growth prediction under grazing conditions.