(here’s my contribution to the discussion/reaction to @pjdambra‘s comment. not to reopen it, but to explain why the scores are perfectly fine with me. It’s become a bit lengthy and technical I found after typing, but anyway.)

Same here. Having become curious by the many high-scores of @asher I decided last week to buy the Angry Birds HD app for Mac. I already had a Mac Pro with 24 inch screen (1920,1200 resolution), but had never played games on it. I still haven’t, but I got myself a top score nonetheless: 150k exactly for Birdday Party 19-8. Here’s the screenshot:
http://angrybir.de/14L0WXw
The score was obtained while i myself was cutting wood and enjoying a bbq afterwards.

The point is that once you have the right strategy, the main/only difficulty is to repeat it exactly, to overcome the randomness built into the Angry Birds levels (each run is different; levels/their elements have random aspects), and repeat until the stars aline. Anything helping in this repetition gives an advantage. The advantage can be huge. For instance, if your eyesight/fingers allow you only an accuracy within 2 pixels on a tablet you will have to spend 25 times more time (expected) than someone being able to repeat pointing at the same pixel, to get the same results, and then i’m not speaking about the timing of the tapping. So yes, I always believed asher’s scores were valid, but I also believed the playing field was not level. Of course, he’s not to blame for that and his scores are very good, even given a precise pointing device. Still, I thought scores could be improved upon further, by improving upon the automation offered by such a pointing device.

Last week I decided to take repetition/automation one step further and this week have made a small Angry Birds playing program. Tried it first at Level 19-8 since that’s a 1-bird strategy (but otherwise the level was chosen randomly), with the above result. (For the technically inclined: the program is written in Java using the Robot class. At the moment it is in proof-of-concept-state (< 200 lines of code); just automating the flinging, with parameters the angle and the tap-delay, making screenshots (the above one was made automatically), some code to facilitate simple strategy programming like reading the score from the screen and seeing whether some elements are present at some location, and a little bit of AI to aid in the exploration/exploitation phases. No physics/trajectory code yet, nor code for automatically recording videos.)

Since Level 19-7 was yesterday's challenge I've now adapted the program to that level (a simple adaptation; 4 lines of code per bird, 12 lines in total for the 3-bird strategy). At this moment it's running and after a few hundred runs its high score is 171610, again beating my own score for that level. In general, I guess not many of the present top scores would survive. (Because of its reactive nature; timing aspects, Programming AB is imo more challenging than programming, say, a chess engine. I had offered this as a project to my AI students, but nobody had taken it thus far.)

I won't enter these scores (all my scores entered here at ABN are my own obtained by pointing with my fingers at an ipad 2 screen), but I am not sure it would be cheating; it treats Angry Birds as a black box as usual, just using its standard input (the event queue) and output (the screen). In fact, advanced gaming mouses already go a long way into this direction/automation of AB, as they allow you to program your mouse buttons to take aim and also to tap automatically after a fixed delay.

Other than lagging a bit behind in its version (e.g. it still has the mad hats in Birdday Party), i found the Mac HD version plays the same as the IPad version i'm used to. It may be that because of the higher resolution some new aiming angles have become available (*) and that therefore in some levels higher scores are possible, but imo the main thing if you want to improve your scores by buying the PC/Mac version is to get yourself a better pointing/clicking device (programmed or not, depending on what game you want to play; e.g. a competition between programs would have my interest.)

(*) To explain, imagine a very-low-resolution screen of 3 by 3 pixels; if the sling would be in the middle, you could only fling at angles which are multiples of 45 degrees. In general only finitely many rational angles will be available. (Wouldn't it be interesting to allow for real instead of just rational coordinates, say, to have an input device allowing to construct a (computable) real coordinate, irrespective of actual screen resolution, by successive refinement?)