Xardion
Explorer
Gonna necro this thread, as I came up with a pretty interesting way to roll stats. But first, some background.
My group (myself, my wife, and some friends) has always preferred rolling stats to using point buy, and while the standard 4d6 drop 1 method works well enough, we always end up needing to roll multiple sets. We've always had a certain expectation of power level from the PCs, and there's enough randomness even in the standard method to end up with either multiple sub-8 scores in a set, or just really bland sets of mostly all 10's or 11's.
So, I sought out a way to alleviate this while still rolling dice (any discussion of point buy in my group will usually get thrown out the window). The guidelines were scores from 7 to 18 (so nothing below a 7 can be possible), and the upper end (16 to 18) needs to have about the same probabilities as the standard method.
I wrote a little program to test various methods and calculate probabilities, and finally stumbled on a perfect solution that does EXACTLY what I set out to do.
The core of the method is rolling 3d4 + 2d3 + 1d6, keep 3 dice, and add a flat 3. Now, if you're doing the math, you'll notice that only allows for scores of 6 to 17, and that the probabilities for 16 and 17 are on the low side compared to 4d6 drop 1. So, there are some special rules added to fix those issues.
First, it's important that you don't use ACTUAL 3-sided dice, and instead are using 6-sided dice to roll the d3's. Second, the actual d6 needs to be a different color, since you'll be treating it as is. The special rules come up when you roll triples on the 6-sided dice (reading all the numbers AS IS, not treating any of them as d3's), or if you roll quads on the d4's and the actual d6 (so you'd obviously not look for quads if you roll a five or six on that die). Here's the triple and quad rules:
The triple and quad rules have multiple effects. First, they preclude the possibility of rolling a 6, and make rolling an 18 possible (a 4+ triple, or a quad four). Also, they shave off some of the probabilities for rolling a 9 or 12 and use that to bring the 16 and 17 probabilities in line with 4d6 drop 1.
After doing a test with 20 million rolls, the probabilities look like this:
I also did some tests with actual dice, and the results were pretty good. None of the sets I rolled would have been considered bad or dull, though a couple sets with end up with 2 16's or 17's, which is expected with the somewhat increased chance of a 16 vs 4d6 drop 1. Didn't roll an 18 on any of them, which I didn't expect to as I only rolled 5 sets. The single set with a 17 was also the only time I rolled a triple or quad (a triple 3), so it doesn't happen that often.
EDIT: Another thing you can do is limit the number of triples and quads in a set, like 1 of each or just 1 of either, which I would recommend doing. It opens the possibility of getting a 6 again, but a 6 really isn't worse than a 7.
My group (myself, my wife, and some friends) has always preferred rolling stats to using point buy, and while the standard 4d6 drop 1 method works well enough, we always end up needing to roll multiple sets. We've always had a certain expectation of power level from the PCs, and there's enough randomness even in the standard method to end up with either multiple sub-8 scores in a set, or just really bland sets of mostly all 10's or 11's.
So, I sought out a way to alleviate this while still rolling dice (any discussion of point buy in my group will usually get thrown out the window). The guidelines were scores from 7 to 18 (so nothing below a 7 can be possible), and the upper end (16 to 18) needs to have about the same probabilities as the standard method.
I wrote a little program to test various methods and calculate probabilities, and finally stumbled on a perfect solution that does EXACTLY what I set out to do.
The core of the method is rolling 3d4 + 2d3 + 1d6, keep 3 dice, and add a flat 3. Now, if you're doing the math, you'll notice that only allows for scores of 6 to 17, and that the probabilities for 16 and 17 are on the low side compared to 4d6 drop 1. So, there are some special rules added to fix those issues.
First, it's important that you don't use ACTUAL 3-sided dice, and instead are using 6-sided dice to roll the d3's. Second, the actual d6 needs to be a different color, since you'll be treating it as is. The special rules come up when you roll triples on the 6-sided dice (reading all the numbers AS IS, not treating any of them as d3's), or if you roll quads on the d4's and the actual d6 (so you'd obviously not look for quads if you roll a five or six on that die). Here's the triple and quad rules:
The triple and quad rules have multiple effects. First, they preclude the possibility of rolling a 6, and make rolling an 18 possible (a 4+ triple, or a quad four). Also, they shave off some of the probabilities for rolling a 9 or 12 and use that to bring the 16 and 17 probabilities in line with 4d6 drop 1.
After doing a test with 20 million rolls, the probabilities look like this:
I also did some tests with actual dice, and the results were pretty good. None of the sets I rolled would have been considered bad or dull, though a couple sets with end up with 2 16's or 17's, which is expected with the somewhat increased chance of a 16 vs 4d6 drop 1. Didn't roll an 18 on any of them, which I didn't expect to as I only rolled 5 sets. The single set with a 17 was also the only time I rolled a triple or quad (a triple 3), so it doesn't happen that often.
EDIT: Another thing you can do is limit the number of triples and quads in a set, like 1 of each or just 1 of either, which I would recommend doing. It opens the possibility of getting a 6 again, but a 6 really isn't worse than a 7.
Last edited:
