And it seems that most of sizes are multiples of 4. Is there any reasons?
Sizes are actually
powers of two: 8, 16, 32, 64, 128, 256 etc. There's historical reasons for this: computers and game consoles use binary data, and the power-of-two numbers are the largest possible number that fits in a binary number of a given size (for instance, an 8-bit console had 256 possible values for a "word" of memory). So in old days, having power-of-two sizes was the most efficient way to pack data, because no space would go to waste. NES cartridges started out with 32
kilobyte of total disk space, so being storage-efficient was crucial. On the NES, tiles were 8x8 in size, which means 64 pixels x 4 colors = 256... i.e., one byte of memory exactly.
I'd guess most modern retro games stuck with it because of tradition... tons of people grew up with 16x16 tile sizes and it's in that sweet spot between "each pixel is a meaningful detail" and "things actually look like the thing they're supposed to be depicting" that makes them easy to work with
and pleasant to look at. Myself, I use 16x16 grids for that reason, but I'm very loose with forcing stuff to be grid-sized. Usually I make 16x24 characters to get more humanlike proportions (with tall characters being 18x29 pixels if needed) but having that set grid size is still really helpful just to know the rough proportions of things before you draw them, so you don't end up with 2-floor ovens and hand-sized cars.