If you want to do something like that, you should look into distributions
I wrote a small bit about distributions in my SmartKeyboard thread (cba to find it lol), but to sort of target the bullseye as it were (and then hit the bullseye 99% of the time), you'd take a normal distribution for both x and y:
Point widgetPoint = new Point(getNormalDistributedRand(x1, x2), getNormalDistributedRand(y1, y2));
Another way is to add up 2 randoms, eg 2 6-sided die have the highest probability of rolling a 7.
int randX = rand(1, 6) + rand(1, 6);
Reading the link I sent again, read how they note "depends on placement". This is where statistics starts to become a very big part of botting.
If you look at my API, I take a lot of this into account. Look at my NPC file here, note the hover(), getSuitablePoints(), and the variety of random functions. What we're essentially saying when we hover is we're trying to match a point that is as close to the mouse as possible (ie doing less work; in words of Fitt's law getting as small an index of difficulty as possible) that is also a suitable point on the NPC. Think of it as a car with thinking and stopping distances. You want the stopping distance to be as small as possible, so we can technically say that we want to stop at the centre, and by the time we enter the rectangle we have finished thinking.
Sounds very confusing, but it's about as human-like as you can get.