If it is possible to estimate fingertip force, you will know the proper fingertip force, such as a cooking, massage, or pet handling expert. Our goal is to establish a method for estimating the force applied to fingertips from muscle deformation sensors attached with a forearm. In previous studies, machine learning-based methods were developed with muscle deformation information for estimating fingertip force. However, the condition of the muscles is changed in several hours, and thus the laborious calibration was required to be performed many times. In addition, since the system was calibrated with only one pressure sensor, it could not handle various pressure applications with multiple fingers, resulting in low accuracy. In this study, we present an estimation method: PondusHand, that automatically calibrates the force of multiple fingertips when pushing the mobile phone panel in various styles each time. In the experiment, it was confirmed that our method could estimate the fingertip force with twice the accuracy of the existing method (error 0. 7N, n=6). In future, our system will be applied to various fingertip force quantifiers, such as pottery, musical performances, cooking, and palpation.