The ability to control redundant motor effectors is one of hallmarks in human motor control, and the topic has been studied extensively over several decades since the initial inquiries proposed by Nicholi Bernstein. However, our understanding of the influence of stroke on the control of redundant motor systems is very limited. This study aimed to investigate the effect of stroke-related constraints on multi-finger force control abilities in a visuomotor task. Impaired (IH) and less-impaired hands (LH) of 19 hemiparetic stroke survivors and 19 age-matched control subjects were examined. Each hand repeatedly produced isometric forces to match a target force of 5 N shown on a computer screen using all four fingers. The hierarchical variability decomposition (HVD) model was used to separate force-matching errors (motor performance) into task-relevant measures (accuracy, steadiness, and reproducibility). Task-irrelevant sources of variability in individual finger force profiles within and between trials (flexibility and multiformity) were also quantified. The IH in the stroke survivors showed deficits in motor performance attributed mainly to lower accuracy and reproducibility as compared to control hands (p < 0.05). The LH in stroke survivors showed lower reproducibility and both hands in stroke also had higher multiformity than the control hands (p < 0.05). The findings from our HVD model suggest that accuracy, reproducibility, and multiformity were mainly impaired during force-matching task in the stroke survivors. The specific motor deficits identified through the HVD model with the new conceptual framework may be considered as critical factors for scientific investigation on stroke and evidence-based rehabilitation of this population.