imho, as there are more qualified people that can better explain this, but since you didnt get a response, i thought i would try and tackle it
it has to do with when the spark occurs in the cylinder... as the motor increases in rpm (or performance capability), ideally (i would think) you would want the point at which the spark occurs to change... i think in our cases, with most of our motors, we are setting it to occur at the best point, when the motor is higher in rpm...
if the spark occurs too soon, it will be in a sense fighting against the piston... it needs to occur when the crank/piston are on the power stroke, just passed top dead center (or i believe right before the piston starts to travel in its power stroke)... if you could imagine the piston being all the way up in the cylinder, close to the head, the the spark occurs at a precise time, igniting the fuel and compressed air, causing the explosion that forces the piston back down and crankshaft to spin...
if the timing is off, say too far advanced, the spark would occur too soon and try and force the piston back down when its on its way up... you would lose power this way. if to far retarded, then i believe the explosion would be caused after the piston was already on its way down, you would lose power this way too... if you could imagine that...
say you are pushing a kid on a swing... you pretty much want to push the kid, right as the swing reaches the point where it is stopping, but before it starts to go back down... imagine pushing the kids back, too early, while the swing is still in full momentum up toward you... and then likewise, after the swing has passed its tdc and already on its way back dowm... you lose momentum either way