This study investigated retrospectively the predictive value of routine semen analysis in pregnancy by in vitro fertilisation (IVF). The selected (n = 796) cycles were divided into two groups: pregnancy group (group 1; n = 264) and nonpregnancy group (group 2; n = 532), in which the female partners were normal or just had tube problems. No significant differences were found in the percentage of normal sperm morphology, sperm motility, sperm progressive motility, rapid progressive motility (rapid) and concentration between the two groups (P > 0.05). However, teratozoospermic index (TZI) and sperm deformity index (SDI) showed statistically significant differences between the two groups (P < 0.05). The number of retrieved eggs (P = 0.001), fertilisation rate (P = 0.000) and number of embryos transferred (P = 0.020) in group 1 were significantly higher than those in group 2, but no significant differences were noted in cleavage rates, and good quality embryo rates between the two groups (P > 0.05). Using receiver operating characteristics curve, we found that semen parameters (morphology, motility and concentration), fertilisation rate, TZI and SDI were not good indicators for pregnancy by IVF. Thus, the semen parameters evaluated according to criteria of the World Health Organization are no good predictors for accurately identifying the IVF outcome. However, TZI and SDI may be more informative than other semen parameters.