For my Pong game I have the paddle follow the ball once it is past the center of the screen. Also if you hit the edge of the padlle the ball goes faster increasing the chance of the computer missing.

the best thing about designing ai player is you can make them do exctly what you want or don''t.
To make the pong player more human, give it a reaction time, so maybe it could start moving in the wrong direction for a sec, or maybe give it a speed attribute that changes with its emotions, e.g. If its lost a lot it will be angry and have a high max speed. (im talking about the paddle there)...

This is a smart and fair AI solution for pong...But naturally, it''s more complex than the ''set the paddle to the vertical location of the ball'' approach : )

This is a somewhat realistic approach as well, since this is what I do when I''m playing pong : )

For starters, each paddle can only move so fast...no jumping the paddle around for the CPU player...

Ok...every ''update'', or every second, or whatever you think a realistic amount of time between judgements is, you look at where the ball is, and where it''s heading. You throw a vector on top of it, which represents where you think it''s going, with some level of variance. You track where the ball goes, and if it hits a wall, you find the opposing vector, with some variance, and when your proposed ball path hits the CPU side, that''s where the paddle tries to move towards, one step at a time. So, if the ball is going to hit 5 times before getting to your paddle, you could be way off on where it''s going to land, but if it''s not going to hit at all, you''ve got a great chance fo knowing where it''s going to land. Lastly, the distance the ball has to travel should also be a factor. Because, when the ball is really close, you''ve always got a good idea of where it''s going.

Difficulty levels? Increase the variance. Impossible, near zero variance (aka, near-perfect prediction).

G''luck,
-Alamar