One of the widely used denoising methods in different domains is the Savitzky-Golay (SG) filter. The SG filter has two design parameters: window length and the filter order. As the length of the window increases, the estimation variance decreases, but the bias error increases at the same time. Mean square error (MSE) measure includes both bias and variance criteria. In this paper, we obtain the optimal window length of an SG filter with arbitrary order which minimizes the MSE. To achieve the optimal window length, we propose an algorithm whose performance is better than the existing methods. In this paper, we follow the viewpoint proposed by Persson and Strang and design the filter on the basis of Chebyshev orthogonal polynomials