Update the top n rows of a table in Postgres

Question

I have the following table:

 year | value  | tercile
------+--------+---------
 1980 |   96.4 |
 1981 |    727 |
 1982 |  804.7 |
 1983 |    979 |
 1984 |    995 |
 1985 |  984.4 |
 1986 |  994.6 |
 1987 |  98.39 |
 1988 |     97 |
 1989 |  87.87 |
 1990 | 4054.3 |
 1991 |  82.89 |
 1992 |     94 |
 1993 |  97.61 |
 1994 |  99.68 |
 1995 |   99.9 |
 1996 |   9.42 |
 1997 |  75.66 |
 1998 |   2.16 |
 1999 |    696 |
 2000 |  99.38 |
 2001 |   9.81 |
 2002 |  99.27 |
 2003 |   2.75 |
 2004 |  12.51 |
 2005 |   9.29 |
 2006 | 994.71 |
 2007 |  97.06 |
 2008 |  64.44 |
 2009 |     95 |
 2010 |  40.88 |
 2011 |   9.46 |
 2012 | 742.86 |
 2013 |    9.5 |
 2014 | 736.38 |
 2015 |  93.13 |

I need to populate the 'tercile' column with values 1-3 according to the values in the 'value' column. 3 is assigned to the records with the 12 highest values, 2 is assigned to the records with the 12 middle values, and 1 is assigned to the records with the lowest 12 values.

I can retrieve these records with the following queries:

select * from trace_terciles order by value desc limit 12 
select * from trace_terciles order by value desc limit 12 offset 12
select * from trace_terciles order by value desc limit 12 offset 24

But I cannot figure out how to update the results of these SELECT statements. How can I do this? By using a subquery?

Answer 1

Update all rows in one query:

update trace_tercile t
set tercile = case
    when rank < 13 then 3
    when rank < 25 then 2
    else 1 end
from (
    select year, rank() over (order by value desc)
    from trace_tercile
    ) s
where s.year = t.year;

Answer 2

How about

update trace_terciles set tercile=2 where year in (select year from trace_terciles order by value desc limit 12 offset 12)

and similarly for the other 2 cases.