I've just had a look at your file Tazinlwfl, there's just a few things I noticed and straigtened out.

The ball that was moving initially is off-center. This is helping to cause an unsatisfactory bounce. I also noticed that this ball was rotated some. A quick reset transformations, before moving the ball back to the rest position solved that.

Next, I found that by placing the axis of a sphere at the point of contact between the frame and the two nails suspending it, caused undesirable results. I simply reset the axis to the center of the sphere.

After I'd straightened out the above, there still remained the question of how to get the balls moving. If the axis was left up between where the 2 nails touched the frame, and the ball rotated, it took the right position but was positively useless during simulation.

What I found was that the ball responded to the dynamics simulation much better if the ball was moved into it's initial position rather than being rotated into it.

So, I created a locator, scaled it to make selection easier and then positioned it at the center of the sphere that was to start the animation. I then placed the axis of this locator between where the two nails that suspend this ball touch the frame. I was then able to rotate the locator into a good position to start the animation(45 deg). Then I grabbed the ball, moved it so it's center was v-snapped to the locator. Since the ball was moved and not rotated, I then had to rotate it by the same amount that the locator was rotated by (45 deg).

If I didn't use a locator to position the ball at it's starting position, I couldn't be sure it was the right distance away from the point it was suspended from - meaning that it'd hang higher or lower than it should have.

Simon

EDIT: wondered about a link to the walk-thru, but see that that's since been covered.