X backlash and Y backlash are both used to compensate the backlash of the mount on both axes. By default, these parameters are ignored.
With high end mounts these parameters can be ignored.
To compensate backlash you should enter the time, in second, that is elapsed without visible motion when the mount axis moves into the opposite direction.
To measure your mount backlash on both axes, you can start the guiding loop without correction and use the guider relay to move the mount to the opposite direction until you see a change in the guiding error. Once you see an error, you should calculate the backlash time by summing the total pulse time used to move and enter the time in the backlash field. You can proceed as this for both axes.
To do a good measurement, you should consider a large move in one direction before starting to go to the opposite direction.
If your mount offers an backlash compensation we suggest using it instead. In any case backlash compensation may no be perfect or may be tricky to get right, some mounts with too much of it may just not be good enough for deep space imaging and should be reserved for visual work. There is a couple of trick that can be done for minimizing the mechanical backlash itself, without any software compensation.
First make a mount East heavy, since the RA always moves in the same direction (no motor reversal) this insure that any gears are always in contact without any play or backlash. In second you could offset a bit your polar alignment to create a systematic DEC drifts.
Unlike for the RA axis, the DEC motor may reverse, however with a proper systematic error from a miss-polar alignment the DEC axis will need a constant, small, correction in the same direction without any motor reversal, hence without backlash issue anymore.
The art of such trick is to offset the polar alignment enough to achieve such systematic DEC error without having to use too short guider exposure an mount correction nor leading to field rotation. Some try and error is likely to be necessary. Usually when backlash is controlled this way the outcome is better than with a software backlash compensation.