A good way to analyze the differences between the two distributions is to imagine a head-to-head contest between characters.
First, suppose you have two identical characters, A
and B
, rolling off against each other with d20. They tie 5% of the time; 47.5% of the time one wins; 47.5% of the time the other wins. In contrast, if you use 3d6, ties occur 9.2% of the time and each wins 45.4% of the time. Not a huge deal. Let's discard the ties and just concentrate on who wins more, A
or B
. Now let's start giving them bonuses. Since we haven't said who is whom, we'll just declare that A
is the stronger one and B
is the weaker one.
A's bonus 3d6 d20 3d6 ratio
========= ===================== ===================== over
========= A-wins B-wins ratio A-wins B-wins ratio d20 ratio
--------- ------ ------ ----- ------ ------ ----- ---------
+0 45.36% 45.36% 1.0 47.50% 47.50% 1.0 1.0
+1 54.64% 36.31% 1.5 52.50% 42.75% 1.2 1.2
+2 63.69% 27.94% 2.3 57.25% 38.25% 1.5 1.5
+3 72.06% 20.58% 3.5 61.75% 34.00% 1.8 1.9
+4 79.42% 14.46% 5.5 66.00% 30.00% 2.2 2.5
+5 85.54% 9.65% 8.9 70.00% 26.25% 2.7 3.3
+6 90.35% 6.08% 14.9 73.75% 22.75% 3.2 4.6
+7 93.92% 3.59% 26.2 77.25% 19.50% 4.0 6.6
+8 96.41% 1.97% 49.0 80.50% 16.50% 4.9 10.0
+9 98.03% 0.99% 99.0 83.50% 13.75% 6.1 16.3
+10 99.01% 0.45% 220.0 86.25% 11.25% 7.7 28.7
+11 99.55% 0.18% 552.9 88.75% 9.00% 9.9 56.1
+12 99.82% 0.06% 1663 91.00% 7.00% 13.0 127.9
+13 99.94% 0.02% 6661 93.00% 5.25% 17.7 376.0
+14 99.98% 0.00% 46649 94.75% 3.75% 25.3 1846.3
Okay, so what does this tell us?
First, we can see that with big bonuses, A
slaughters B
head-to-head in rolls in 3d6, whereas with d20 the benefit that A
gets over B
is pretty modest (has to get all the way up to +11 before A
is tenfold more likely to win than B
!).
But, second, if you look at the ratio of ratios (that is, how much advantage A
vs B
has in 3d6 compared to A
vs B
in d20), you find that in 3d6
the bonus is pretty much squared compared to d20 (low values only--then it gets way, way more extreme later on).
So, what does this mean? Well, basically, if under 3d6 you have a +1 bonus more than someone else, it feels like a +2 difference in d20. +7 feels like +14.
So the concise explanation is: moving from d20 to 3d6 amplifies differences, making them feel about twice as large as before. (Of course, almost nothing is actually resolved as a head-to-head test, but it's a useful thought experiment.) You can cleave through hordes of lesser beings with that much more ease, and your betters become that much more fearsome. In fact, better just stay away from them. There are some kobolds that need slaying. Right? Right.
Compiled Results from Other Answers
DnD Next numbers include calculations from both the 1st and 2nd playtests.
Fighter Rogue Wizard Sturdy Wizard
OD&D 11 3 2 -
AD&D 14 6 2 -
3.5 11 6 3 4
4e(MM1) 13 10 7 9
4e(MM3) 11 8 6 7
Next test1 10 8 6 -
Next test2 12 4 3 -
Summary
AD&D: improved fighter & rogue survivability
3.5: slightly improved wizard survivability, and pulled fighter survivability down considerably (trend towards narrower range begins)
4e: improved everyone's survivability, though mostly rogue & wizard, further narrowing the spread
4e's MM3: reduced all survivability and tightened the spread again
DnD Next (playtest 1): slight reduction in fighter survivability to tighten the spread even more
DnD Next (playtest 2): major reversal of the reduced spread trend
Thoughts
From 3.5 on, every edition change (including the switch inside 4e from MM1 damage expressions to MM3 damage expressions) has essentially worked to reduce the survivability gap between the toughest and weakest PCs, primarily by bringing the fighter down but in 4e's case by bringing the wizard up. We're down to fighters lasting about twice as long as wizards, rather than the 5-7 times longer from OD&D and AD&D. Rogues have moved from being only marginally more durable than wizards to being about halfway between wizards and fighters. Base wizard survivability has approximately tripled since OD&D/AD&D, and later editions have given them more options for improving it further.
A Note on HP Inflation
As of playtest 1, worries about hit point inflation in D&D Next over 4e appear unfounded: D&D Next PCs last about as long as 4e PCs do when using the new 4e monster damage values, and only slightly less than 4e PCs do when using the original 4e monster damage values.
As of playtest 2, DnD Next hit points are back to pre-3rd standards.
Best Answer
Don't do it.
The XP guidance in the DMG already cover how party size factors into things.
If your goal is...
...you're doing exactly the opposite.
The encounter system is designed with the standard array in mind. Deliberately weakening characters across the board is going to require you to rebalance everything by eye, rather than using the guidance in the book.