Too many when all are thrown
In the first section I threw everything but the counts proved too big. In the second section I threw only hearts, stars, and wheels, but the wheels rolled creating extreme outliers. The hearts, stars, and letters work best.
I throw the beads and then count the number in each row of tiles on the floor. The number rises and falls rather normally about the impact point of the beads. The beads have a small amount of bounce which helps with the dispersion. The tiles are twelve inch tiles, which seems to work best. Probably anything could actually be thrown such as paperclips, but the height might have to be increased.
The point is that the normal distribution occurs purely randomly and naturally. Any time stuff is dumped onto the floor, there is basically a normal distribution to the mess. I use this to also explain why statisticians spend so much time with the normal distribution - the normal distribution is what one normally encounters. Sure there are lots of other good looking distributions - binomial, Poisson, and many others. But lots of things can be done with the good old normal curve.