So far, we have accomplished real time flying baseball image extraction from filmed pictures and real time three-dimensional coordinates calculation using stereo-vision systems [1][2][3]. And also we have proposed baseball arriving point forecast algorithm using thus obtained ball three-dimensional coordinates. This paper proposes two algorithms to improve measurement accuracy. Forecast accuracy is directly related to real-time observed ball coordinates accuracy. To get higher forecast accuracy, it is effective to increase the number of ball coordinates observed. First algorithm is to double effective TV frame rate. NTSC type TV images are divided into every field to double observed ball coordinates numbers. Second one is to get an algorithm to give the robot catcher the best control timing for catching the ball, and forecast arriving points. The effectiveness of those algorithms has partially been confirmed experimentally.