I think with your helpful analysis I am drawn to saying that cards and points-buy are the same for MAD. It might be that the card mix should be tuned further, but I am not sure how to approach the probability analysis. For instance, are those arrays with 15s that are serving One and Two, more or less probable than the array with two 14s?
Well, I am glad to help and find it interesting. Point-buy is perfect for MAD classes as you can build a character with three good abilities and no penalties. You can draw a three score decent MAD character with cards, but because it is balanced out, you will have two penalties.
As far as those three arrays, the single 15 (one) is the most likely, two is next, and three with the 14's is least likely. This is because of how the draws have to happen. For a single 15, you only need one score to draw 3 of the 5's. For array "two", the fives are required as well as one of the 4's. But for 2 14's in three, four of the 5's and all of the fours have to be in those scores. So, even with the card system, it is hard to get MAD class scores supported if you want them to be higher scores. If you are happy with a more balanced, but decent scores, such as {12, 12, 12, 10, 9, 8} that is not hard to get and prior to racial adjustments would be perfectly acceptable to me.
I continue to suspect that 4d6k3 will usually put one and two ability classes further ahead of three than the more constrained systems. Compare the likelihood of one super-score to that of three. Sometimes the 4d6k3 MAD character will be fine or, very occasionally, amazing.
Actually, no, 4d6k3 is more likely to have more "better" scores because it has replacement. In the card method, if you get a 15, those 5's are not available later and consequently your scores will be lower. When rolling 4d6k3, if I roll a 15, all the numbers are still available later on for me to roll another.
If you recall a while ago, this was one reason why I asked if your idea was with replacement or not. With replacement in 4d6k3, I could get an array of all 13's, all scores above average of 12.45. With cards, I can't get an array of all 11's, with all scores above the average of 10.5.
FYI, here are the probabilities for the individual scores as there are 816 combinations you can draw for a single result. Later draws are dependent on earlier ones, but ultimately the probabilities remain consistent because if you draw certain numbers first, others are available later, and vice versa.
Score | N | Prob. |
6 | 10 | 0.0123 |
7 | 40 | 0.0490 |
8 | 70 | 0.0858 |
9 | 134 | 0.1642 |
10 | 154 | 0.1887 |
11 | 154 | 0.1887 |
12 | 134 | 0.1642 |
13 | 70 | 0.0858 |
14 | 40 | 0.0490 |
15 | 10 | 0.0123 |
TOTAL = | 816 | |