I would recommend doing the calculation manually:
select avg(case when col > t.avg then col - t.avgcol end),
avg(case when col < t.avg then t.avgcol - col end)
from table t cross join
(select avg(col) as avgcol) as tavg;
This gives the average. If you want the "variance", just square the differences, sum them and take the square root:
select sum(case when col > t.avg then (col - t.avgcol) * (col - t.avgcol) end) / sum(col > t.avg),
sum(case when col < t.avg then (col - t.avgcol) * (col - t.avgcol) end) / sum(col > t.avg)
from table t cross join
(select avg(col) as avgcol) as tavg;
I suspect that you are learning to apply these concepts on real data. You might be interested in learning about statistical skew which provides another measure of how "centered" the average is. A good place to start is Wikipedia.