hive query case when status "case when , rand"

Question

this:

drop table if exists temp_a;
create table temp_a
as
select 
 case when rand(123) < 0.4 then 1 
      when  rand(123) >= 0.4 and rand(123) < 0.8 then 2 
      else 3 end as label 
from source_data ;

select label, count(1) as count from temp_a group by label;

but the result is :

label   count(1)    
1       111175     
2       80509       
3       87690      

distribution 
label  count / sum
1      40%
2      28%
3      32%

it does not like 40% 40% 20%, why?

i want to know "why the distribution not like 40% 40% 20%"