Edit: Oops forgot to address your point too, which yeah you're right, there are ways to get around bad distributions from poor generation like what you proposed, and it would work to smooth it out and not feel so crappy to the player for sure. --------- The reason why is because newer programmers will naively initialize their random generator, and/or use a standard random generator rather than a more robust (and more random) one, resulting in rolls something like this: > rand(10) > 1 > 1 > 1 > 1 > 6 > 9 > 9 > 9 And you might say, "Well that's a perfectly random result", which it could be when utilizing a hardware random generator with actual entropy, but CPU based random generation is only pseudo-random, so the initialization and choice of algorithm can affect the results such that clustering of same results in rather unlikely distributions like that become more common. Often times when RNG in a game seems bad it's simply because it is.