The Chi-Chi earthquake that occurred in central west Taiwan is one of the most well-recorded large earthquakes in the world. Consequently, one should be able to reliably locate the point of nucleation from the impulsive onsets on short-period seismograms. But in fact, the hypocentral locations reported by various agencies are different. Our approach is to relocate the Chi-Chi mainshock with the most unambiguous arrival readings and a 3-D velocity structure using the Master Station (MS) method. In addition to the usage of a reasonable 3-D velocity model, the MS method is superior in two aspects: it is unaffected by the nonlinear nature of traditional earthquake locating process and does not have the trade-off problem between the focal depth and the assumed origin time. Our results indicate that the best solution of hypocenter always converges to 23.86±0.01°N, 120.81±0.01°E, and a depth of 10 km, no matter which initial conditions are adapted. This location is essentially the same as that later refined by CWB using a combined bataset of both arrivals on short-period seismograms and S-P time intervals measured from near-source strong-motion records. An important implication of our study is that reliable hypocenter can be derived using as few as 10 accurately determined arrival times from nearby short-period stations when a reasonable 3-D velocity model and the MS method are used in the hypocenter determination. Thus, the performance of CWB RTD system, which is already the leading example of its kind, can be further improved in the near future.