Looks like a LOWESS curve (https://en.wikipedia.org/wiki/Local_regression). They always overfit, but are still useful to show trends. The main danger is that they get wonky near the limits of the data. Note that that increase at the end of the left plot for Democrats looks like it is increasing -but that increase looks to be 100% dependent on that single data point for '25. Obviously, you never want your analysis to be dependent on a single point like that.