Great article!

Just out of curiosity, I took a child that achieved a perfect score and ran its DNA through the scoring procedure a few more times. To my surprise, the results were all over the map (ranging from -335 to 490). This led me to try removing the robot's ability to make random moves. Because the robots now perform consistently, no top score in any generation was lower than the generations that came before it, which was not the case in the original example.

Thanks for making my brain work today!