Although the average magnetopause is ~10 RE from the Earth, the magnetopause moves inside the geosynchronous orbit during extreme solar wind conditions. Under these circumstances some geosynchronous satellites suddenly enter the magnetosheath and are exposed to the plasma and fields of the magnetosheath. In this study we evaluate the predictive capabilities of magnetopause location models in forecasting geosynchronous magnetopause crossings. We predict periods during which geosynchronous satellites enter the magnetosheath using the Petrinec and Russell [1996> and Shue et al. [1998> magnetopause location models driven by data from Interplanetary Monitor Platform (IMP) 8. These predictions are then verified with in situ observations from Geosynchronous Operational Environment Satellite (GOES) 5, 6, and 7. We estimate the false alarm rate, probability of detection, and probability of false prediction for the two models. The estimation shows that false alarm rate for a forecast with a 20-min separation cadence is ~62% (80%) for the Shue et al. [1998> model (the Petrinec and Russell [1996> model). The probability of detection is very high for both prediction models. These results suggest that both models work well in predicting magnetosheath periods for geosynchronous satellites. Predictions from the models provide a prerequisite condition for geosynchronous magnetopause crossings. Further examination of unsuccessful events indicates that preconditioning by the interplanetary magnetic field Bz needs to be included in the forecasting procedure for a better forecast. This finding provides us with a guide to improving future magnetopause location models. ¿ 2000 American Geophysical Union