SQL Server 2017

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.
create table temp 
(
    id int identity(1,1),
    a decimal(6,2),
    b decimal(6,2),
    c decimal(6,2),
    d decimal(6,2),
    e decimal(6,2),
    f decimal(6,2),
    g decimal(6,2),
    h decimal(6,2),
    i decimal(6,2),
    j decimal(6,2),
    k decimal(6,2),
    l decimal(6,2),
    m decimal(6,2),
    n decimal(6,2),
    o decimal(6,2),
    p decimal(6,2),
    q decimal(6,2),
    r decimal(6,2),
    s decimal(6,2),
    t decimal(6,2),
    u decimal(6,2)
) ;
insert into temp
    (a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u)
values
    (1,5,6,7,8,2,6,3,4,5,2,1,6,5,7,8,2,7,6,2,8),
    (1,5,6,7,8,2,2,3,2,4,2,1,4,5,9,8,2,7,6,2,8) ;
2 rows affected

select id, avg(val)
from ( 
    select id, val
         , count(*) over (partition by id) as c
         , row_number() over (partition by id order by val) as rn
    from temp unpivot (
             val for col in (a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u)
         ) as x 
) as y
where rn IN ((c + 1)/2, (c + 2)/2) 
group by id; 


id
(No column name)


1
5.000000

2
4.000000


select t.id, z.val
from temp as t
  cross apply
    ( select top (1) y.val
      from 
        ( select top (11) x.val
          from
            ( values (a),(b),(c),(d),(e),(f),(g),(h),(i),(j),
                     (k),(l),(m),(n),(o),(p),(q),(r),(s),(t),(u)
            ) as x (val)
          order by x.val
        ) as y
      order by y.val desc
    ) as z ;


id
val


1
5.00

2
4.00