Compression involves recognizing repeated patterns and storing them in a non-redundant way in order to reduce the file size. There are few basic compression algorithms to do them.
Ziv-Lempel is one of them I was taught at my IT courses.
I am attaching a file that I just created. it is almost 4GB, but when compressed it is just about 1.8MB . File contains a repeated single character. File size is high since it is containing RAR dictionaries etc.

file.rar
- (1.82 MiB) Downloaded 771 times
(Can't zip this file, it is too large)
Those ripped games use very high compressions. And decoding them is going to be extremely resource consuming.
And I have no idea how the OS image in your given link is compressed so much. Although I *irated this copy of windows, I got it from a place I trust.