I haven't checked out this thread in a while. Just an interesting experience to add to the mix.
Just a reminder, my preference is to roll stats, in order. I also like a little bit lower average point value than 4d6k3 provides, but also want to minimize characters with very low point value.
So, we roll 3d6, reroll 1's once, in order. This provides an average of about 69 points instead of 72 of the RAW standard array.
If the stats rolled don't work for you, you can swap any two scores.
If you don't want to roll, you use our standard array (14, 13, 12, 11, 10, 9) which also totals 69 points.
One of the fallacies that has been oft-repeated here is that rolling produces overpowered characters. The reality is that it doesn't, but since many DMs allow a player to reroll a "bad" character, it does skew higher since they discard the poor ones. So our rules were tweaked to avoid that.
In our case, we roll 3 characters, and all three are completed as characters. We also randomly determine some innate or unusual talents, from proficiency in a skill, to something bigger like Lucky (which is a lesser version of, and replaces the feat), to things like darkvision that comes from an unknown ancestry. Only about 50% of the characters will have any of these or, rarely, more than one.
The purpose of our approach is to encourage creativity by working with "what you're born with." All of the random stuff is completed first, and then they select the race and class from there. I also have a reduced number of races and classes to choose in my campaign.
What I didn't expect, since this current group includes a lot of new players, is that they all preferred rolling, regardless of the math involved. Specifically rolling the stats in order. The reason being is that it made it much easier for them to quickly create a character, and it narrowed their options to make initial decision-making much easier.
I will say, that for one new player that had three specific characters in mind (and states outright that he's a power gamer), we made sure those would work within the rolls (including bumping up a couple of scores), and nobody had any issue with that approach. We just took what was rolled, rearranged, and added a couple of points to fix one of them.
The rest of the group still want to continue rolling randomly and creating characters for every set, good or bad, because they loved the characters they came up with. And almost every one of them picked as their "favorite" character at the start as one of the less optimal, if not the least optimal. The general commentary is that the wider spread makes it easier for them to develop an idea about who that character is.
The campaign is one that will utilize all of the characters who are rolled. We've played a couple of sessions with the first group, but one of the players won't be available for at least a week, so we're switching to a different group this week. I think that really goes a long way to ensuring that nobody feels they are stuck with a bad character, but it also encourages using all of them at one point or another to give them a chance.
Don't get me wrong, I'm not saying this is better than other methods, but it was very interesting to see how a new group has reacted to my general approach. I did tweak the method and the math a bit from what I learned in this thread, and I think it's working even better than other versions I've used in the past. The players in this particular campaign come from a variety of gaming backgrounds, including AD&D, 3, 4, and 5e, board gamers, MtG, video gamers, and about half of them completely new to D&D. It's definitely the approach that's right for us.