The baseline is indeed the average temperature for a pre-defined interval. However, it is the average for the dataset being considered. In other words, you produce the baseline average after you’ve done the adjustment. The anomalies are then relative to that baseline. Therefore, it doesn’t matter if you shift the buoys up to the ships, or the ships down to the buoys. Once you’ve done the shift, you then calculate the average for the pre-defined interval and present your data as anomalies relative to that baseline.