We present a multidisciplinary approach for predicting how sperm cells with various morphologies swim in three-dimensions (3D), over time scales of milliseconds to hours at spatial resolutions of less than half a micron. We created the sperm 3D geometry and built a numerical mechanical model using the experimentally acquired dynamic 3D refractive index profiles of sperm cells swimming freely in vitro as imaged by high-resolution optical diffraction tomography. By controlling parameters in the model, such as the size and shape of the sperm head and tail, we can then predict how different sperm cells, normal or abnormal, would swim in 3D, in the short or long term. We quantified various 3D structural factor effects on the sperm long-term motility. We found that some abnormal sperm cells swim faster than normal sperm cells, in contrast to the commonly-used sperm selection assumption during IVF, according to which sperm cells should mainly be chosen based on their progressive motion. We established a new tool for sperm analysis and male-infertility diagnosis, as well as new sperm selection criteria for fertility treatments.