
CAN ENTROPY DECREASE FULL
Now, if you happen to have full information about all of the universe with precise state of every particle, you can still reverse it. The corase entropy has increased, and the process is irreversible. As they do, the phase space volume occupied by these states increases. Small, tiny changes in original states of particles in the glass or floor make huge differences in where the cracks head and how the glass flies into shards. A glass falling on the way to the floor is a good example. The important part is that if the system exhibits no chaos, and all trajectories flow together, then the coarse entropy never changes. There is a way to state this with mathematical precision, but I don't think it will be useful. It's the quantitiy that doesn't just concern itself with total number of states, but rather the rough volume these states occupy. Precisely, this is described with coarse entropy.

And yet, the details of the impact and how the glass shatters, will depend on everything that's happening. If I drop a glass on the floor, I do not need to know anything other than composition of the glass and maybe small region of the floor near impact to predict what happens. What's far more interesting is considering a small corner of the universe, which, nonetheless, interacts with the rest of the universe. It's an object that just is throughout all of the time, and any dynamics within it is fully reversible. It's boring, just like the universe as a whole is boring.

There is the fine grain entropy, which is defined for the universe as a whole, and it is a conserved quantity. Loosly speaking, there are two types of entropy we consider. This is why typical explanations of entropy, these not amed at people who spent many years studying mechanics, chaos, thermodynamics, and field theory, well, frankly, suck. If entropy increases only in one direction (from "past" to "future"), this contradicts "time reversibility".
