Imagine you have two boxes. In each of them you have a perfect, frictionless rubber ball, labelled A and B. You've shaken the box with A inside a lot, and the ball is bouncing around inside it... high energy! The other, B, is just sitting at the bottom of its box, low energy.

Now you put the boxes together and remove the wall between them. What happens? Soon the bouncing ball hits the stationary one. It's a glancing blow, but B now has a tiny bit of energy, and A has a little less. The total energy is the same. Soon, it happens again! This time a lot of energy transfers. Now B has a little more than half the energy, and A a little less than half!

As you watch, this keeps happening. The balls keep trading energy between them. Sometimes A gets a bit more of the energy, sometimes B, most of the time it's about even. It IS possible for all the energy to go back to A... but the balls have to hit JUST RIGHT for that, and there are far more ways they can hit where that doesn't happen.

Now repeat the experiment with 1000 rubber balls in each box. Again the ones in box A start with all the energy. The same thing happens, when the wall goes down the A balls slam into the B balls and everything just reaches an equilibrium quickly. Sure, sometimes one rubber ball gets a huge kick, maybe because two others slam into it at once, but on average there isn't really a difference between the balls labelled A and the ones labelled B anymore. It's even more unlikely to spontaneously arrange itself into a state where the A balls have all of the energy again -- ALL 2000 balls would have to collide perfectly at once for that to happen. But the total energy remains the same.

That's "what decides", on the microscale -- chance. On the macroscale, the resultant statistics. Things don't really tend to low energy, they tend to equilibrate. If you throw a hot rock into a pot of cold water, the water gets hotter and the rock gets colder, but the water doesn't get hotter than the rock (on average) and the rock doesn't get colder than the water (on average)... they come to be the same temperature... and if you have a perfectly insulated pot, the total energy inside doesn't change even if the water and rock are changing.