This is a tough one for improved properties, because there are usually so many other variables to waterfront properties (highly custom homes, differences in quality of water, view, etc...). It is difficult to isolate just the frontage component. But, my first move for something like this is to identify a group of somewhat similar properties (maybe your comps, possibly including older sales you are not using as comps). But, I try to find 5-8 data points. Put them on a grid and adjust for all other differences including time. If other adjustments are accurate, then all remaining price variance can be attributed to the frontage. Create a scatter plot with the partially adjusted price on the Y axis and frontage on the X axis. Put a trend line on the chart, which will hopefully show a clear relationship between frontage and price. The trend line equation can be used to make the individual adjustments. This works if the trend line 'fits' the data well. There are nuances to this approach, and it won't work if the data is inconsistent or just too messy, but this is a limiting factor for any adjustment extraction method.