The winsorized mean is a statistical measure of central tendency that is similar to the mean and median. It involves replacing extreme values at the high and low ends of a distribution with the most extreme remaining values. Typically, around 10 to 25 percent of the ends are replaced. The winsorized mean is considered a robust estimator because it allows for the retention of outliers without being overly influenced by them. However, it may not produce unbiased estimators for the mean or median unless the underlying distribution is symmetric. Overall, the winsorized mean is a useful tool for generating reasonable estimates of central tendency in statistical models.
https://en.wikipedia.org/wiki/Winsorized_mean