Most baseline problems in instrumental methods are characterized by a smooth baseline and a superimposed signal that carries the analytical information: a se-ries of peaks that are either all positive or all negative. We combine a smoother with asymmetric weighting of deviations from the (smooth) trend get an effective baseline estimator. It is easy to use, fast and keeps the analytical peak signal in-tact. No prior information about peak shapes or baseline (polynomial) is needed by the method. The performance is illustrated by simulation and applications to real data.