From the results of the three local smoothing methods, the window span is very important among the factors that can affect smoothing. If the window size is smaller, the loess method is better fitting the height fluctuation model. For functions of Mexican fat, it is important to use a lower window span. The goal is not to use too many data points in the fit to avoid excessive bias. If the data used is overfitted, we first need to ensure that the model is smooth. However, deviations from the original function can occur. If the fitting aspect is considered separately, NW method performs best in this aspect. But the disadvantage of the NW method is the high variance between the Monte Carlo and the average values. Finally, it is important to adjust the parameters of the local smoothing method. The performance inversion of loess and spline smoothing methods shows that the bias, variance and mean square error can vary greatly and need to be optimized by tuning parameters