This shell script does what you want, but there is probably a better way to solve your problem.

Code:

#!/bin/sh

while read line; do
	c1=$(echo "$line" | awk '{print $1}')
	c2=$(echo "$line" | awk '{print $2}')
	c3=$(echo "$line" | awk '{print $3}')
	c4=$(echo "$line" | awk '{print $4}')
	if [ "$c4" = '+' ]; then j=$(($c2-$c3)); else j=$(($c3-$c2)); fi
	[ ! "$p" ] && p="$c1"
	if [ "$p" != "$c1" ]; then echo "$this"; i=""; p=""; fi
	[ -z "$i" ] && i="$j"
	if [ "$j" -le "$i" ]; then i="$j"; this="$line"; fi
done < file
echo "$this"

exit 0

The datafile is called file in this script
From c1... to c4..., we register each field from the datafile file in a variable.
We then check if c4 is a + or - and act accordingly.
p is a variable to remember the last c1, on the first pass, it does not exist, so p = c1
if p != c1 (if we change from x to y or z), we display the line this that we set further in the script and we reset i (used later) and p.
if i does not exist (which is the case on the first pass), we set it to j, calculated above
if j is less than or equal to i, then we set j to i, and we keep the corresponding line from the datafile in the variable this
We go back to the top and check for the next line in the file.