Good news for all pen & paper roleplayers: We aren't completely crazy, all the myths about some dice being luckier than others are true. Dice with rounded edges are indeed not perfect, and certain numbers can thus appear up to 20% more or 20% less than average, because of how the dice are "tumbled" to round the edges. Nevertheless the claims of Gamescience that their dice with sharp edges are superior because giving a result closer to randomness have hit a snag: Due to these dice having a chunk of plastic sticking out on one face, the opposite face of that dice comes up 40% less than average. On the Gamescience d20 that is the 14 which is rarer than others.

For practical purposes all these dice are sufficiently random for a game. The rounded edge dice, like those from Chessex, have the advantage of their deviation from randomness being random, thus you can have "lucky" or "unlucky" dice. I don't like the idea of a Gamescience dice being predictably non-random, even if 14 is not a particularly significant number on a d20 in most games. Thus if you wanted your dice to be as random as possible, you would have to remove the uneven spot from where the plastic is injected in the mold. But of course if you don't do it perfectly, you risk another deviation from randomness.