Baking (at least for games) is where you "transfer" the high resolution mesh details to the optimized game mesh. The high poly mesh can have all the detail you want: bumps, creases, pores, insets, among a whole host of things. The optimized game mesh is much less detailed, following the critical parts of the silhouette. No bumps, no scratches, just the bare minimum. Baking then allows the detail from the high resolution mesh to be transferred to an image, which a real-time rendering engine can load with less impact on the resources than if you were to use the high poly mesh. There are different things you can bake to help give the illusion of a higher resolution asset but still have it run in real-time.
I hope that makes sense and helps.
No prob! You can also bake out certain animations, meaning instead of having the computer do the calculations in real-time (such as with particles or simulation), you do the precalculation, then save the calculations out as an animation. So, it's kind of like doing mental math and using a calculator.
Hope that helps, too.
I presume if you "bake" it then you have simulation ready before render, so less render time .... my opinion i am n00b :D