I'm doing something that I don't think has been done very often. I'm writing a character generator for a tabletop RPG. (Please note, that first sentence was irony.)

I'm considering two methods for rolling stats.

1. Roll the eight stats each time.
2. Include a two dimensional array, something like 10000x8 and only generate one random number from one to the length of the array. Whichever number is generated, take the eight values from that row. One advantage of this method is not worrying about a duplicate seed if I use srand(time(null)) to seed.

First, I'd like your thoughts on the two methods.

Second, I'd like your views on the efficiency of doing it either way.

Thanks,
Doug

Well, to be quite frank, "it has been done to death." However, there is always room for one more . . .

"Efficiency" notwithstanding (it absolutely does not matter ....), the question is simply whether-or-not you wish for your pseudo-random results to be biased:

• If you do not want a bias, simply "roll the dice."
• If, for the purposes of your game, you do wish to introduce bias, allocate a temporary array that will serve as "a deck of (sorted ...) cards," with whatever bias you wish to have, pre-programmed into the distribution of that deck. Then, select a card at random from that deck.

"Efficiency" is an irrelevant consideration. The statistical distribution of possible outcomes is the only criterion that applies, and all strategies for getting there are equally plausible.

In truth, I want to have it so the GM can decide about the bias. The game uses 3d6 to generate the stat. I'm going to give him the ability to set whether it's straight 3 to 18, 9 to 18, or a high power game, 12 to 18.