I think rather not. You will quite probably need a group by to get the sum per day and those result then matching with one or two peek-loadings (with different sortings to get the previous records from both sites) and afterwards a further aggregation to get the average.
Alternative to the mentioned peek-runs could I think on a kind of As-of-Table within the script (each day is there matched with the 10 days prior and after it) and which will be taken to flag the right days together. A different way might be to match the days simply to their RowNo in 20-buckets (but it will be difficult if there gaps between them).
In conclusion I think there would be alternatives possible but I'm not sure if they would be more elegant or better perform than the "classical" way.