Looking for a way to better do this

sift.pl

#sifting through a list,  the keys are non unique and can show up any number of times with corresponding #level; aqua,purple,blue
#The desire is to sift each key to its highest value; aqua > purple > blue.  
#I am using a sub like this to handle updating the value in a hash that uses the key value to 
#uniquify the data and will then give me the key with it's highest rated level.  
#This seemed a pretty efficient way to do it but I'd like input.  
#It's perl but really more interested in the algorithm.   


sub sift_value{
    my $value_in = shift;
    my $value_old = shift;
    return $value_in if $value_in =~ /aqua/i;
    return $value_in if $value_in =~ /purple/i && $value_old !~ /aqua/i;
    return $value_old;
}