Near-infrared (NIR) spectroscopy combined with the moving window partial least-squares (MWPLS) method and Savitzky–Golay (SG) smoothing was successfully applied for the rapid no-reagent determination of chemical oxygen demand (COD) in sugar refinery wastewater. An appropriate waveband with stability was selected through a rigorous modeling process. Among 141 samples, 60 were randomly selected as the validation set. The remaining 81 samples were divided into the calibration set (50 samples) and the prediction set (31 samples) for a total of 20 times with certain similarities. The results showed that the short-wave NIR region (780 nm to 1100 nm) can be used as the information waveband of COD in sugar refinery wastewater, and the optimal SG smoothing mode was 5th order derivative, 5th degree polynomial, and 7 smoothing points. The waveband selection was performed in the SG smoothing spectra of the short-wave NIR region by the MWPLS method. The optimized waveband was 808 nm to 964 nm, the validation root mean square error of prediction (V-SEP) and validation correlation coefficients of prediction (V-R_P) were 25.0 mg L⁻¹ and 0.966, respectively, which had good prediction precision and stability. The results provide a reliable NIR analysis model and valuable references for designing small, specialized NIR instruments.