Though violin has been widely used in classical music, violin performance involves a problem that how to determine fingering from a musical score. In fingering determination, beginners often consider the ease for playing, but intermediates take account of expressiveness as well according to their skill level grown. Therefore, we aim to develop an automatic fingering estimation method to assist both beginners and intermediates in finding their appropriate fingering. The problem of fingering estimation is basically to find a most probable state sequence, where a state represents a pattern of left hand corresponding to a musical note from the score. We define the state of left hand as a combination of four parameters. The appropriateness of states is locally defined as a state transition cost and pressing string cost. The problem is how to set the model parameters to adjust the desired skill level. We focused on that the note length affects the difference in violin fingering. In our method, appropriate fingering according to the skill level can be estimated by changing relation between the note length and expressiveness. By comparing estimated fingering with textbook's fingering and subjective evaluation, we confirmed that our method can estimate appropriate fingering for a desired skill level.