m20 stats, part 3

Yeah, yeah... I'm noting that my m20 stat generation methods are less interesting than my ramblings about design decisions and playstyle preferences in D&D.  But that's where my RPG head is for the moment.  I think I've narrowed in on how I want to do this, although I've thought of a few more radical options.  Here's some general principles:

• Random generation feels more "natural" to me.  I like rolling up stats better than point buy for visceral reasons.  The problem, of course, is that random stats are... well, random.  It's nice to say that you want randomized stat generation, but it generally sucks if the dice treat you poorly.  Since stats are something that you live with the entire campaign, or at least until your character dies and needs to be replaced, it usually isn't very fun to have the character who rolled up really crappy stats, and while it can be an occasionally fun challenge to play Barney Fife as a D&D character, that's not really the reason most of us come to the hobby.
• It's all well and good to tell someone to suck it up and live with what they rolled, but let's remember here: this is our hobby.  Within reason, we should be doing it to have fun.
• I'm not as concerned about this, but it is a concern nonetheless--the guy who rolls really poorly and the guy who rolls really well in the same campaign can perform very differently in game.  This can lead to further imbalances of play.  The biggest single concern here is the player who doesn't get his moment to shine, and so never even really enjoys the campaign very much, but there are other issues as well--especially if the campaign is meant to be fairly difficult, and there are characters that can't pull their own weight.  Even if it's not meant to be difficult, the notion of why an adventuring party--a group of highly skilled and mutually compatible experts in their chosen fields--would lug along such obviously dead weight.
• The easy solution to this is point buy, which is--let's face it--quite popular, even in D&D.  It's often the only way to create characters in many other games.  I like it as an option.  But it feels a little too sterile for my tastes.  I somehow want the best of both worlds; characters that have an almost guaranteed skew towards the higher end of average, but with stats that aren't completely bland.  I like some risk, or at least some variability.

There's several ways to do this, with varying efficacy.  d20, for instance, instead of having 3d6 in order, has 4d6, drop the lowest (so you have an effective 3d6, but skewed slightly higher than average) and arranged to taste.  Frankly, after all the years I've done that, even that seems like small potatoes towards skewing.

I have decided to go with the narrower "Moldvay range" of stat bonuses, -3 to +3 instead of the d20 range of -4 to +4.  Of course, that can change once racial bonuses are added, and after a few leveling ups have led to a stat increase or two, but for the initial step, the challenge is to turn a minimum of -3 in all three stats to something more usuable, up to +3, in at least some stats.  You need at least 18 points to get from an all minimum to an all maximum, and to get an average +0 in all stats, you need at least 9 points.  To get a higher than average, but not maximum, you need something there in the middle as a median.

A good way to get a skew into the top half is to only roll for the top half, and let the bottom half (or whatever number it is) be flat.  If, for example, I had players roll stats as 2d4 + 10, then your range of points to spend on bringing your three -3's up would be 12-18.  (This is a bit too good, by the way.  The average would be about 15.  I prefer one closer to 13 or so.) 

So, although it sounds a little weird, I'd like to go with stat generation of either 3d4 + 6 or 2d6 + 6.  The former gives a range of 9-18, with an average of about 13.5 with a bigger distribution around the average, and the latter gives a range of 8-18 with an average of 13 but a slightly more widely disbursed distribution (i.e., slightly more likely to be on the higher and lower end compared to the other method.)

The optional alternative would be point-buy with 13 points to spend.