We present a model-based systematic study of antitachycardia pacing protocols applied to atrial fibrillation, focusing on the ability to achieve and maintain capture during pacing, as a function of both pacing site and period. We observed that pacing sites located away from anatomical obstacles led to faster and more robust capture. Moreover, after comparing burst and ramp pacing, our results indicate that in order to get capture it is necessary to pace at a fixed optimal period over a sufficient long time.