Can someone help me use anydice.com for this problem

[Ashtagon]

http://anydice.com/

I'm looking at designing an alternate character generation system for a homebrew rules system.

The final is to replace the 3-18 stat range with 2-12, and get rid of the odd-number gaps inherent in the 3.x system ability scores. However, I also want to keep the same general probabilities.

Yes, I am aware that any 2dx-based approach will create a wedge-shape rather than a bell curve. I can live with that, as long as the overall average is right.

3d6b4 typically results in a +1 ability score modifier, so for my 2-12-based system, I had considered 3d6b2. However, that results in a typical score of 9 (a +2 modifier in what I hope to be the final system).

So, what I am considering is:

1. 2d6, six times (roll for each of six ability scores).
2. Re-roll any three individual dice.
3. Assign each of the six result to the ability of your choice.

How can I calculate the final probabilities?

[DedWards]

I wouldn't mind trying to help, but could you please explain to me what the b"2" part of "3d6b2" means?

[Amechra]

Take the best 2 dice.

---

I do have to ask, have you considered not having randomly-rolled attributes? Like, going with a point-buy, therefore giving you (the designer) greater control over the range of ability scores that a given starting character has.

Anydice doesn't "do" choice, so you wouldn't be able to anydice it. If you want a rough idea of the probabilities, character creation would look like any of the following, depending on whether or not the player uses the discretionary rerolls:

2d6 6 times.
3d6b2 once, 2d6 5 times.
3d6b2 twice, 2d6 4 times.
3d6b2 thrice, 2d6 thrice times.
4d6b2 once, 2d6 5 times.
4d6b2 once, 3d6b2 once, and 2d6 4 times.

2d6 produces an average of 7, 3d6b2 produces an average of 9, and 4d6b2 produces an average of 10.