We examine the risk model in sperm competition games for cases where female fertility increases significantly with sperm numbers (sperm limitation). Without sperm competition, sperm allocation increases with sperm limitation. We define ‘average risk’ as the probability q that females in the population mate twice, and ‘perceived risk’ as the information males gain about the sperm competition probability with individual females. If males obtain no information from individual females, sperm numbers increase with q unless sperm limitation is high and one of the two competing ejaculates is strongly disfavoured. If males can distinguish between virgin and mated females, greater sperm allocation to virgins is favoured by high sperm limitation, high q, and by the second male's ejaculate being disfavoured. With high sperm limitation, sperm allocation to virgins increases and to mated females decreases with q at high q levels. With perfect information about female mating pattern, sperm allocation (i) to virgins that will mate again exceeds that to mated females and to virgins that will mate only once, (ii) to virgins that mate only once exceeds that for mated females if q is high and there is high second male disadvantage and (iii) to each type of female can decrease with q if sperm limitation is high, although the average allocation increases at least across low q levels. In general, higher sperm allocation to virgins is favoured by: strong disadvantage to the second ejaculate, high sperm limitation, high average risk and increased information (perceived risk). These conditions may apply in a few species, especially spiders.