The aim of this study was to estimate the between days test-retest reliability of nonlinear metrics used to quantify motor variability in a repetitive precision task. On three separate days, 14 healthy subjects performed pipetting as a general model of repetitive precision tasks. The task consisted of transferring liquid 20 times with a cycle time of 2.8 s from a pickup tube to eight target tubes placed on a table in front of the subjects. The motion of hand, arm and the pipet tip was tracked in 3D and the shoulder elevation and elbow flexion angle were obtained. Motor variability was assessed using nonlinear metrics based on information theory and recurrence quantification analysis. Between- and within- (between-days) subject variance components were computed using a one-way random effect model, and intra-class correlation coefficients (ICC) were calculated from the variance components as standardized measures of reliability. Most of the metrics displayed a considerable between-days variance component and therefore the ICC showed a slight to moderate reliability. The reported data on between- and within-subject variability can be used to design future studies using non-linear motor variability metrics on kinematics data.