I think that dropping the pencils and comparing how many - or any, resulting dots, line segments, zig-zags, cusps, and curves on different pieces of paper might invovle modelling a chaotic system. I'm wondering if this model might be comparable to one that could be used to represent a a modeified double pendulum such as the one at Double Pendulum on Flickr - Photo Sharing!. Instead of a pendulum with at least one light that was consistently on, the light could flash. And instead of a double pendulum, there could be an n pendulum, n being equal to the number of pencils that were dropped each time divided by the number of times that the lights on the pendulum model flashed. The models could involve comparing many papers on which many pencils were dropped in one case, and many representations of sequences of flashing lights on the pendulum in the other case. So, in the both models for each trial the number of chaning dots, line segments, zig-zags, cusps, and/or curves may be about the same at any one time and in total for the extent of the trial. I thought about modeling these by dividing the area into a grid, counting events in each grid, measuring distances between recurring events, and counting the number of times between repeating selected sequences in each grid. However, I still think that I'm leaving a lot out and haven't found the best modeling technique for these.