When you toss a coin, you place it in a certain starting position on your fingers, you apply a certain force with your thumb to the particular edge of the coin, it starts rotating within a specific air pressure and movement, only to fall on a surface with a given elasticity and friction. An algorithm, knowing these and other necessary parameters could determine the outcome before the coin leaves your fingers.
When you throw a dice, you put it in your fingers in a certain way, apply certain force with your arm to give it speed, spin your fingers to give it rotation, then it bumps and rolls on the specific surface, movement determined by dice center of mass and the friction and other surface properties only to stop with a given number facing up. A machine with accurate enough visual sensors, capable of measuring the forces involved, the surface of operation and the dice build, would determine the outcome of the throw before the dice touches the table.
When there’s a car accident, it’s an effect of many decisions of many living things on and off the road leading to that one mistake that causes the crash. Maybe it was a running cat that lead that first car to break, and the driver behind it to break even harder in panic, and you changing the shitty radio station to something classy at that moment, which lead to hitting the breaks late and ramming the back of that poor panicked driver. The cat is gone, so is the first car, but had you know… But a powerful machine could know — having data on everything from traffic movement, animal tracking and your music taste it would predict this situation. Could it prevent it? Probably not without altering the scenario and predicting something else. But it would know and it’d no longer be random.
The truth is, nothing’s random. We just don’t know how to show it.