Eye gaze accuracy is one of the most commonly used parameter to evaluate the eye tracker performance. The current study applied stepwise multiple regression to predict the significant predictors for eye gaze accuracy (AC). 7 male and 3 female were recruited to perform multi-directional tapping task in the projection-based stereoscopic display under 3 different levels of parallax and 6 different levels of index of difficulty (ID). Tobii X2 eye tracker was used to measure the selected four eye movement parameters which consist number of fixation (NF), fixation duration (FD), time to first fixation (TFF), and eye movement time (EMT). The results indicated that NF was found to be the best predictor for AC followed by EMT and parallax. The R2 value of 0.247 indicating that the 24.7% of the variability of the data was explained by the model. Practitioner Summary: The result of multiple regression can be a valuable theoretical foundation for evaluating an eye tracker. The results could be very beneficial for human factors engineers and virtual reality developers especially for predicting eye gaze accuracy.