Coded as such:

use strict;
use warnings;

# Initialize
my ($x, $y, $z) = (-11,-13,4);
my $maxodd;        # Intentionally left as undef

# Check $x (though $maxodd will be undef, write it consistently in cas
+e code is revectored later)
if ($x % 2) {
    if ( (!defined $maxodd) || ( (defined $maxodd) && ($x > $maxodd) )
+ ) {
        $maxodd = $x;
    }
}
# Check $y
if ($y % 2) {
    if ( (!defined $maxodd) || ( (defined $maxodd) && ($y > $maxodd) )
+ ) {
        $maxodd = $y;
    }
}
# Check $z
if ($z % 2) {
    if ( (!defined $maxodd) || ( (defined $maxodd) && ($z > $maxodd) )
+ ) {
        $maxodd = $z;
    }
}
# Report results
if (!defined $maxodd) {
    print "All are even numbers\n";
} else {
    print "$maxodd is the greatest odd number\n";
}

# Fini
exit;
[download]

This screams for the use of a subroutine, though I suspect the author isn't at that point in the book yet.