This paper discusses new criterion for the evaluation of load frequency control (LFC) performance. The proposed LFC performance evaluation is based on a standard deviational ellipse that characterizes a bivariate relationship between the frequency deviation and the area's interchange power variation. The standard deviational ellipse is evaluated using Tyler's maximum-likelihood estimator, which is distribution-free within the family of elliptical distributions. In addition, a target standard deviational ellipse is proposed, using predefined values for a target frequency bound and the area's frequency bias. The presented approach is based on the accordance of the estimated ellipse with the target ellipse, resulting in several grades for the LFC performance. A comparison of the proposed criterion is made with other LFC performance criteria known from the literature. Extensive measurements were applied, and the obtained results show the correctness of the proposed approach.