Ability Score Generation

Recommendation for Basic D&D: Ability Score Generation Methods



Default Standard Array Method: 15, 14, 13, 12, 11, 10



Optional Point-Buy Method: 26-Point Buy

Ability   Point
Score:  Cost
(18)       (50 points)
17:         30 points
16:         26 points
15:         12 points
14:           8 points
13:           3 points
12:           2 points
11:           1 point
10:           0 points
  9:         −1 points
  8:         −2 points
  7:         −3 points
  6:         −4 points

When the points of all six ability scores total 26 points, the effectiveness of the array will be about as effective as the Standard Array.

Examples of arrays generated by the 26-Point Buy:

• 17 10 10 10 10 6
• 16 12 10 10 10 8
• 15 15 12 10 10 10
15 14 13 12 11 10
• 14 14 14 12 10 10
• 14 14 13 13 12 12



Optional Dice-Roll Method: Roll 2d6 + 4 for each score for three arrays, then choose any array.



Note about methods: Each player at the table may choose any of these methods to determine the array for the player character. The methods should generate comparable arrays. So dont worry too much what the other players are getting. Now, the Dice-Roll Method is a gamble and can generate a noticeably better or worse array. But the gap between it and the other arrays should be narrow enough to allow all characters to perform comparably.

Wizards, shave and a haircut

The current Playtest (01 28 13) offers three methods of generating scores for the abilities. Here are the arrays that each method generates

• Default: 4d6 Method ≈ (16,14,13,12,11,10)
• Optional: Array Method = (15,14,13,12,11,10)
• Optional: Point-Buy Method ≈ (15,15,15,8,8,8)



The problem is, the default method of Roll 4d6 Drop Lowest is statistically superior to either optional method.

The Playtest default of the 4d6 Method generates an average array that, for the 4e Point-Buy Method is equivalent to about 22 or 21 points (where 22 is the 4e standard), and for the 3e Point-Buy Method is equivalent to about 30 or 29 points.

By contrast, Playtest option of the Array is only worth 20 points for the 4e Point-Buy System, or 28 points for the 3e Point-Buy System. The array option inferior to the 4d6 default by 2 points. Not only that, the 4d6 default is likely to grant a score of 16, and even a 17 or 18 is possible, while alternative Playtest options make 16, 17, and 18 impossible to achieve. These highest scores are extremely valuable, and their presence or absence significantly skews character effectiveness.

Finally, the Playtest Point-Buy Method is equivalent to only a 17 points in 4e, or 24 points in 3e. The Playtest Point-Buy Method is drastically inferior to the other two Playtest methods. The Playtest Point-Buy Method also makes the game-changing values of 16, 17, 18 impossible.



In comparison, the three methods create characters at significantly different levels of power. Worse, because the optional methods are strictly inferior, the incentive is to use the default 4d6 Method. That is the worst of the problem. Because the 4d6 Method is a poor method that is extremely unreliable.

Altho the 4d6 Method generates an average array equivalent to about 30 points in 4e, it actually varies considerably. A group that use the 4d6 Method is like to have one player with an array worth 35 points while an other player suffers an array 25 points. Easily there can be spread of 10 points between player characters. This creates severe gaps in the effectiveness between player characters. With the luck of a single session of dice rolls, one character will dominate with an extreme advantage for the entire champaign, possibly lasting many years in reallife. While an other character suffers weakly.



The main concern is, D&D Next uses “bounded accuracy” as the foundation of the gaming design. Because of this, characters with lower scores are better for the wellbeing of the game overall.

Therefore, the Playtest Array Option (15,14,13,12,11,10) are the scores that need to become the Standard for D&D Next.

The wonky 4d6 Method can be removed from the game, and replaced with a more reliable dice roll method.

Finally, the Playtest Point-Buy Method needs tweaking to generate values that are more comparable to the Array Method.
 
These arrays are equivalent to the Playtest Standard Array. The values of the arrays depend on calculations in an other thread, Arrays.

Compare quickly with a set of arrays that cost a different price. This “Elite Array”, 15 14 13 12 11 10, costs 12.5 points.

Here are arrays that cost 12.5 points:

17 11 10 10 10 6
16 12 10 10 10 8
15 14 13 12 11 10
14 14 14 13 11 10



Because the Playtest Array Method uses 15 14 13 12 11 10 as the Standard Array. The values above are other arrays of comparable value.


These are the values that the Playtest Point-Buy Method should calibrate with, and that the Playtest dice-rolling method should tend toward.
Asymmetrical pointbuys need to cease to be.  It shouldn't cost (over) twice as much to get to +2, than what going to -2 returns.

Consider this method:
community.wizards.com/go/thread/view/758...
Assymetrical pointbuys need to cease to be.  It should't cost twice as much to get a +2 than going to -2 returns.

Consider this method:
community.wizards.com/go/thread/view/758...


Sorry, but I have to disagree.  Asymmetry is the only thing that even remotely preserves a point buy system that lets you sell back from baseline for points.  This is because a +2 in your prime stat is more valuable than a -2 in a stat you don't regularly use is damning.

"Enjoy your screams, Sarpadia - they will soon be muffled beneath snow and ice."
On Worldbuilding - On Crafting Aliens - Pillars of Art and Flavor - Simulationism, Narritivism, and Gamism - Shub-Niggurath in D&D
THE COALITION WAR GAME -Phyrexian Chief Praetor
Round 1: (4-1-2, 1 kill)
Round 2: (16-8-2, 4 kills)
Round 3: (18-9-2, 1 kill)
Round 4: (22-10-0, 2 kills)
Round 5: (56-16-3, 9 kills)
Round 6: (8-7-1)

Last Edited by Ralph on blank, 1920

It's the DMs job to shut down blatant minmaxing, not the system itself.

Quite frankly, if the system rolls over and dies when a 20/20/20/1/1/1 character somehow shows up, it's just fundamentally flawed.
Here is a dice rolling method. It produces scores generally comparable to the Standard Array. Occasionally a player might get lucky with an awesome array, or unlucky with a blah one. But the spread between the effectiveness of their arrays wont be too terrible. They can probably play their characters at the same table without noticing too much that one is better than the other.



Roll 2d6+4 for each score for three arrays, then choose any array.



Each score will range from 6 to 16. However the lower scores rarely happen, unless you choose them because that array comes with several higher scores.



So far, the method is producing values that I myself can live with if everyone else is using the array. It even is producing values that I could live with if I got one of the blah arrays while an other player got one of the awesome arrays.
Roll 2d6+4 for each score for three arrays, then choose the best array.

That's interesting.
I'd change "best" to "any".  The "best array" is very subjective.
Roll 2d6+4 for each score for three arrays, then choose the best array.

That's interesting.
I'd change "best" to "any".  The "best array" is very subjective.



Fair enough. It really does depend on the character concept. Will update the wording to any.
Rolling stats is for players like me who want to let chaos have a hand in character creation. I for one don't believe every character in a party should be equal, that's boring to me. I like working with the strengths and weaknesses I get, not custom-tailoring my character to be perfect. That's why I'll always roll my stats and have my players do the same.

But the other options should be offered of course for every player's satisfaction, and they should be balanced to each other. I do see the problem with the standard array being too low. I think it should have a 16, because in many years of rolling 4d6 almost all of my characters have had a 16 or higher.

Obviously there is no way to actually balance a radomized system with a non-randomized system, but 16,14,13,12,10,8 would be my suggestion, not based on math but observation from many randomly generated characters.
By the way, I updated the Recommended Point-Buy System in the Second Post. The math is a bit tighter, and this way of transitioning to whole numbers looks cleaner. Let me know, if these point costs generate arrays that you feel are moreorless comparable to having the Standard Array.



Rolling stats is for players like me who want to let chaos have a hand in character creation. I for one don't believe every character in a party should be equal, that's boring to me. I like working with the strengths and weaknesses I get, not custom-tailoring my character to be perfect. That's why I'll always roll my stats and have my players do the same.

But the other options should be offered of course for every player's satisfaction, and they should be balanced to each other. I do see the problem with the standard array being too low. I think it should have a 16, because in many years of rolling 4d6 almost all of my characters have had a 16 or higher.

Obviously there is no way to actually balance a radomized system with a non-randomized system, but 16,14,13,12,10,8 would be my suggestion, not based on math but observation from many randomly generated characters.



I understand the suggestion for your higher array. Certainly, in previous editions, like 3e and 4e, your array is probably the minimum that I could be happy with.

But D&D Next is using bounded accuracy, so starting with lower numbers works better for the gaming system, as the adventure advances upto Level 20. Given the characteristics of the score values, the bonus values, the d20 increments, and so on, it is a wiser foundation for the math of the game to start with the Standard Array: 15 14 13 12 11 10. Also the symmetry of this array is appealing. It even looks like it makes sense.

Dont forget, both the race and the class grant an additional boost this score, turning the 15 into a 17, the 14 into a 16, or so on. So the character starts with scores that are far above average.

For D&D Next, I have come to love the Standard Array. It makes an excellent standard, so it is the dice-roll method and the point-buy method that need to recalibrate with this Array in mind.
Problem is, recalibrating the dice roll method would be next to impossible. All you can really do is go down to 3d6, which drastically lowers results, not slightly. I don't see any reason to mess with a system that's been in every edition of the game without problems.

Another problem with the current standard array is that only Humans are capable of starting with an 18 in a stat. Other people have complained about this. With a 16 in the array, any race can have an 18 to start with, and humans can have a 19, but they both can have a +4.

That might be too much with bounded accuracy, but Humans will get it either way, so at least they shouldn't be the only ones. Unless we wanted to greatly lower the starting array and make 3d6 the standard roll, or if we weakened Humans and didn't change the array or the roll.
Problem is, recalibrating the dice roll method would be next to impossible. All you can really do is go down to 3d6, which drastically lowers results, not slightly. I don't see any reason to mess with a system that's been in every edition of the game without problems.


Actually, the dice roll method I came up with for this thread, now in the Original Post, seems to be working amazingly well. I ran the method about 200 times so far, and every single time it has produced an ability array that seems fair, comparing to the Standard Array. Only once did I get a score that was eyebrow-raising too good, and once that was probably too mediocre. But even these outliers seem workable at the table, and make the gamble interesting. As far as I can tell, I am happy to use this method for my own characters, when Im in the mood for a gamble, and happy too if an other player uses it for their characters.



Another problem with the current standard array is that only Humans are capable of starting with an 18 in a stat. Other people have complained about this. With a 16 in the array, any race can have an 18 to start with, and humans can have a 19, but they both can have a +4.

That might be too much with bounded accuracy, but Humans will get it either way, so at least they shouldn't be the only ones. Unless we wanted to greatly lower the starting array and make 3d6 the standard roll, or if we weakened Humans and didn't change the array or the roll.


I empathize. Mike said, they will get around to revamping the racial features. When they do, I (and many others) hope the Human ability boosts will make more sense compared to the other races.
I'm working on an “Advanced Option”, that will generate arrays old-school, in the style of AD&D.

Once upon a time, D&D used to roll 3d6 to determine an ability score. (Not 4d6!) The array that this method generates is game-breaking. There is a reason the D&D tradition mainly abandoned it. Therefore this isnt an option that a player can choose on their own for their character. It requires a group decision where everyone is on board. Moreover, the player who takes on the responsibilities of DMing must be an expert in the rules of D&D, and know how to improvise to accommodate both the player that performs exceeding powerfully and the one exceedingly poorly.

That said, there are ways to mitigate the 3d6 method to make it work better in the context of D&D Next.

The 2e Players Handbook lists several official options for how to generates charcters.

I dislike the option of roll 4d6, dropping the lowest d6, for each score of one array. This 2e method is what 3e adopted. The problem is, the 4d6 method is too likely to roll a gaming-breakingly good character. At the same time, it is likely to roll a mediocre character for one of the other players at the table. There is no mechanical way to mitigate the result, and the need to override it subjectively (almost always because the result is disappointingly bad) has created a systemic culture of cheating. The need to override the result ruins the point of rolling randomly in the first place.

Fortunately, 2e offers other array methods. Of interest here is this method: Roll 3d6 for each score for two arrays, then choose any array. This method is still harsh and disruptive. However, if we increase the number of arrays from two to three, the results are a bit more stable. The ability to choose from three arrays, helps the player avoid getting stuck with a fatally bad character. Also because the method actually rolls 3d6 (not 4d6), it is extremely unlikely to roll an 18. The method tends to avoid brokenly good characters. Bad things can still happen - because it is 3d6 after all - but much less frequently.

Moreover, the math of 3d6 for Three Arrays works well with the math of D&D Next. Altho the resulting arrays are wild and unpredictable, the average corresponds closely with the Standard Array (15 14 13 12 11 10).

Notice this 3d6 for Three Arrays is mathematically similar to the method that this thread recommends: 2d6+4 for Three Arrays.

This Advanced Option is for players who want the high-risk gamble when creating their character. Old School players will recognize the return to a pure 3d6 roll for an ability.
For the sake of a reality check, I used both methods to generate 50 characters, each.

The methodology is intentionally intuitive. It boils down to, Can I live with this character? More specifically, if I had the choice of going with the Standard Array (15 14 13 12 11 10), do I now regret getting stuck with this random result?





I rate each array method by how it compares to the Standard Array.

X: broken good (absurd, such as an 18 with 17s and 16s)
a: too good (noticeably better than the Standard Array)
b: ideal (ideal for gaming math, comparable to the Standard Array)
c: disappointing (noticeably worse than the Standard Array)
F: fail (below average and with at least three values below 8)



2d6+4 Method (Recommended for Basic D&D)

Roll 2d6+4 for each score for three arrays, then choose any array.



Results:

bbcbb cbaba bcabb abcbb bbbab bbcbb cbabc bbcbb bcabc abcbb

X: 0%
a: 16%
b: 62%
c: 22%
F: 0%



The 2d6+4 Method allows players in a Basic game to “mix it up” with regard to their ability scores, without worrying about gameplay or gaming math.

(b) As you can see, roughly two thirds of the resulting arrays are comparable to the Standard Array.
(a,c) Note, even the arrays that are noticeably better or worse tend to remain not to far away.
(X,F) The method produces zero failed characters.




3d6 Method (Old School) (Recommended for Advanced D&D)

Roll 3d6 for each score for three arrays, then choose any array.



Results:

Fbbcb baacc abbFF aXaba abacc abbbc abccb cacaa acbca baabb

X: 2%
a: 34%
b: 32%
c: 24%
F: 6%



This 3d6 Method averages toward the Standard Array, thus at least tending toward the math at work in D&D Next. Nevertheless, the method is more likely than not to deviate away from the ideal value. 

(b) Roughly one third of the results generate characters that are ideal.
(a,c) Roughly two thirds of the characters are either noticeably better or noticeably worse than the Standard Array. Even so, the gamble seems somewhat fair, even leaning slightly better than the ideal. Because the player chooses the entire array, it often includes one or more scores above 14 or below 8. Even a single high score can make the entire array more appealing. Even tho (a,c) lists as simply as better or worse than the Standard Array, the 3d6 Method is much more wild than the 2d6+4 Method. For example, the 3d6 Method will often include a low score with a significant penalty.
(X,F) Roughly a tenth of scores are broken.



That said, the old-school style 3d6 Method works better than might be expected. To some degree, it avoids fatal arrays that plague the 3d6 tradition. Because it is extremely difficult to roll an 18, the method also tends to avoid the upper extremes. 





3d6 x10, keep the middle six results.
3d6 x10, keep the middle six results.



Really?



Are these arrays what you are suggesting? Of the first 10 arrays that this method generates, only two are comparable to the Standard Array, and one of them only marginally. The remaining eight are noticeably worse than the Standard Array.

 

b (comparable to Standard Array)
16, 12, 11, 11, 11, 11
14, 14, 12, 9, 9, 8

c (worse than Standard Array)
13, 13, 12, 9, 8, 7
13, 12, 12, 11, 10, 8
13, 11, 11, 10, 8, 7
12, 11, 10, 9, 8, 8
12, 10, 9, 9, 8, 8
11, 11, 9, 9, 9, 8
11, 10, 8, 8, 8, 8
11, 10, 9, 9, 8, 7



Did I misunderstand you (roll 10×, remove the two highest and two lowest)? This method seems unappealing.

4d6 Method (Not Recommended)

Here is the trial of the 4d6 Method: roll 4d6, dropping the lowest, for each score for one array.

X: broken array
a: better than the Standard Array - in this case often much better, such as with all three 17 16 15 ...
b: comparable to the Standard Array - ideal
c: worse than the Standard Array
F: terribad


Results:

baaab abbaa baXbX bXXaa baaaa abbaa aaaab aaaaX bbabc baaaa
caaba acaaa cbbcb Xaaab acaba aabba bbbaa baaaa baaaa abbaa

X: 6%
a: 58%
b: 30%
c: 6%
F: 0%

When using the 4d6 Method, 64% of the results are better than the Standard Array. Often the 4d6 Method is much, much, better. This means the distance can be extreme between the abilities of one character and the abilities of an other character despite being at the same table.



The 4d6 Method deviates considerably from, and fails to be comparable with, the Standard Array Method (15 14 13 12 11 10).
I like the 2d6+4 idea in spirit, but then I went to roll up a sample block and got stuck with 11, 11, 10, 13, 7, 10. The next two tests were better, but in each case the standard array was preferable. Bad luck, I guess.

Playing a character is a commitment. Those attributes inform just about everything you do in the game. If my goal as a DM at the table is to make everyone sure everyone has a good time, I can't leave that decision up to a roll of the dice. Too many people (myself included) get cranky when their character under-performs. It's no fun to be a rogue with 14-15 dexterity while the wizard miraculously rolled a pair of 16s for int and dex. Jealousy is petty but real.

I agree the 4d6 method is terrible and 2d6+4 looks quite a bit better to generate 5e stats, but I will stick with point buy.
I like the 2d6+4 idea in spirit, but then I went to roll up a sample block and got stuck with 11, 11, 10, 13, 7, 10. The next two tests were better, but in each case the standard array was preferable. Bad luck, I guess.


Heh, there does seem to be that 22% chance for your initial roll (13, 11, 11, 10, 10, 7). I call it a “blah” array.

Just now I did 10 arrays cold. These results are a bit on the low side - the (a) catagory is a bit marginal. Note, I rolled the 2d6 180 times for this method, so statistically I should have come across about five 16s, but I didnt roll any. Even so the statistical distribution seems robust. Notice, one of the arrays that I rolled is the Standard Array itself. Personally, I am fine with playing any of these characters.



(a)
15, 15, 14, 13, 9, 8
15, 15, 14, 12, 11, 8

(b)
15, 14, 13, 12, 11, 10
15, 14, 13, 11, 9, 9
15, 13, 12, 11, 10, 10
15, 13, 11, 11, 11, 9
14, 14, 13, 12, 12, 9
14, 12, 11, 10, 10, 7

(c)
13, 13, 11, 10, 10, 8
12, 11, 11, 11, 10, 9





Even with the batch being on the low side, I am happy to use any of these arrays. For me, the value is, the dice method allows me to get a “fresh” character, without worrying about brokenness, whether too good or too bad.


Playing a character is a commitment. Those attributes inform just about everything you do in the game. If my goal as a DM at the table is to make everyone sure everyone has a good time, I can't leave that decision up to a roll of the dice. Too many people (myself included) get cranky when their character under-performs. It's no fun to be a rogue with 14-15 dexterity while the wizard miraculously rolled a pair of 16s for int and dex. Jealousy is petty but real.

I agree the 4d6 method is terrible and 2d6+4 looks quite a bit better to generate 5e stats, but I will stick with point buy.


You describe alot my sentiment too.

Also, it isnt just envy that is the problem. If the distance between characters is too great, it becomes mathematically difficult for the DM to create exciting challenges. There isnt much room in the gaming math between autowin and autofail. Because of the d20, there is only 9 points between them. If one player has an ability with a +5 bonus, then only 4 points out of 20 are in doubt.
When we look at this ...

(a)
15, 15, 14, 13, 9, 8
15, 15, 14, 12, 11, 8

(b)
15, 14, 13, 12, 11, 10
15, 14, 13, 11, 9, 9
15, 13, 12, 11, 10, 10
15, 13, 11, 11, 11, 9
14, 14, 13, 12, 12, 9
14, 12, 11, 10, 10, 7

(c)
13, 13, 11, 10, 10, 8
12, 11, 11, 11, 10, 9



... what we need to see is this ...

(a)
+2, +2, +2, +1, −1, −1
+2, +2, +2, +1, +0, −1

(b)
+2, +2, +1, +1, +0, +0
+2, +2, +1, +0, −1, −1
+2, +1, +1, +0, +0, +0
+2, +1, +0, +0, +0, −1
+2, +2, +1, +1, +1, −1
+2, +1, +0, +0, +0, −2

(c)
+1, +1, +0, +0, +0, −1
+1, +0, +0, +0, +0, −1
Thank you Haldrik for all this data, it is going to influence the way I allow characters to be rolled in the future.

If they dont want to use the sensible option of point buy or array, I will allow 2d6+4 as a compromise
In my next campaign I'm going to use a communal dice rolling method. For each stat we'll roll a number of times as players and after each roll the players can decide who it goes to.

This will provide the diversity of rolling, allow the group to disperse the extreme rolls so that there's not as much disperity between players, and allow those player that have builds in mind to work towards them.
I am still holding out hope that WotC undoes their prior simplification that really complicated things - by which I refer to their decision to make ability modifiers +/- 1 for every 2 points away from 10.

It made remember what modifier each score had a touch easier, but made rolling your ability scores much more difficult to accurately balance the game around.

Below, I have included a summary of the way ability score bonuses were staggered in D&D (AD&D has similar ranges, but much more granular results) prior to 3rd Edition, and the statistical probability of each modifier if using the then-standard 3d6 method.

3 -3
4-5 -2
6-8 -1
9-12 no modifier
13-15 +1
16-17 +2
18 +3

Easy (enough) to remember because the middle 4 numbers have no modifier, then 3 numbers have +1 or -1, then 2 numbers have +2 or -2, then only one number has +3 or -3.

+/-3 Modifier 0.46% chance
+/-2 Modifier 4.17% chance
+/-1 Modifier 21.29% chance
0 modifier 48.14% chance

Chance of negative modifier 25.92%, with same chance of positive modifier.

Then compare to the same die method with the 3rd editon (and forward) ability modifiers

+/-4 Modifier 0.46% chance, a widening of bonus range being an inherent difficulty in balance calculation.
+/-3 Modifier 4.17% chance, a roughly 900% increase in likelyhood.
+/-2 Modifier 11.57% chance, a roughly 277% increase in likelyhood.
+/-1 Modifier 21.29% chance, the only range that doesn't change.
0 Modifier 25% chance, a roughly 50% decrease in likelyhood.

Chance of negative modifier 37.5%, a roughly 144% increase in likelyhood, and the same for chance of positive modifier.

Used to be the modifiers were based specifically around the idea that most scores would be in the "no modifier' range, so the game math could assume that you didn't have any modifiers - and then each plus was a genuine bonus and each minus a geniune obstacle, but not all that likely to be majorly significant.

...then they decided simple calculation of modifier was more important than being able to set their "benchmark" at a +0 modifier, and decided (if you look at the game math) that they would instead set the benchmark so high as to make people genuinely feel they had to max out their prime ability score or face "uselessness".

With D&D Next aiming for a bounded accuracy system that stands up to the stress of accomodating multipl play (and character generation) styles, I'd have figured putting ability modifiers back the way they were would have been the first step.

ATTENTION:  If while reading my post you find yourself thinking "Either this guy is being sarcastic, or he is an idiot," do please assume that I am an idiot. It makes reading your replies more entertaining. If, however, you find yourself hoping that I am not being even remotely serious then you are very likely correct as I find irreverence and being ridiculous to be relaxing.

It's applying a bell curve to a bell curve.
@Aaron. I can support the increase of bonuses, by every three score points:

Heroic Human
• −1 (7-9)
• +0 (10-12)
• +1 (13-15)
• +2 (16-18)

Superhuman
• +3 (19-21)



But so far, D&D Next seems to solve the problem of disruptively high bonuses by narrowing scores, generally from 10-15. That seems a good solution to me too.
Because of the general uselessness of odd scores, I would be fine if, even-number scores only give the bonus to Checks and Defense, while odd-number scores only give the bonus to attacks and spell DCs.

(If splitting a bonus by three score points: Checks, then Defense, then Attack.)
There are two mechanisms that need to be reconciled here. The first is that some people desire randomness and the second is that some people desire stability amongst players. In order to manage both  you need to create a system that has a autocorrelating factor than diminishes that likelihood of good rolls after already rolling a good roll or bad rolls after already rolling a bad roll. There are ways to do this while still maintaining random distribtuions of stats, but it would be very complicated to explain to a player why they have to go through so much trouble just to simulate a pointsbought array.
Show
I am White/Black
I am White/Black
Take The Magic Dual Colour Test - Beta today!
Created with Rum and Monkey's Personality Test Generator.
I'm both orderly and selfish. I act mostly for my own benefit, but I respect and help my community - Specially when it helps me. At best, I'm loyal and dedicated; at worst, I'm elitist and shrewd.
What Clan are you? There is no equality. There exists only equity.
Why not randomly select from legal point-bought arrays? There's several ways to do this:

1. Create a bunch of arrays using the point buy rules and then have people roll to select one

2. Use a computer program that spits out legal arrays randomly (Wizards could easily make an app that is available for use at the table)

3. Use a hybrid roll/point buy method where you start out rolling for scores but keep track of how much they would cost using point buy (any real low rolls would be bumped up to the point buy minimum); after the first four stats or so you use the remaining points to buy the last stats
Um...I'm not sure how much this affects the math in this thread, but the current ability score array is 15/14/13/12/10/8, not 15/14/13/12/11/10, which corresponds exactly to what is possible in the 27 point point-buying option.
Because of the general uselessness of odd scores, I would be fine if, even-number scores only give the bonus to Checks and Defense, while odd-number scores only give the bonus to attacks and spell DCs.

(If splitting a bonus by three score points: Checks, then Defense, then Attack.)



Now THAT is a cool new idea! You would need to get both at 20 though so you dont end up with a max of +4 to attack/spell DC. That would put an N in next
Problem is, recalibrating the dice roll method would be next to impossible. All you can really do is go down to 3d6, which drastically lowers results, not slightly. I don't see any reason to mess with a system that's been in every edition of the game without problems.

Another problem with the current standard array is that only Humans are capable of starting with an 18 in a stat. Other people have complained about this. With a 16 in the array, any race can have an 18 to start with, and humans can have a 19, but they both can have a +4.

That might be too much with bounded accuracy, but Humans will get it either way, so at least they shouldn't be the only ones. Unless we wanted to greatly lower the starting array and make 3d6 the standard roll, or if we weakened Humans and didn't change the array or the roll.



IIRC, 1st edition, and maybe BECMI, used 3d6, not 4d6, but 4d6 drop the lowest quickly became an option because it lessened the possible variance.

I tried having my players use the 3d6 method for stat generation in the campaign I'm starting, in an attempt to keep their scores within the bounds of what they'd get using point-buy, but the results were less than satisfactory.  Even allowing one of them to reroll three times still resulted in abysmal scores.  I wasn't happy with the results of letting them roll 4d6, either, as the results were too high, but at least they weren't as far off from point-buy.
IIRC, 1st edition, and maybe BECMI, used 3d6, not 4d6, but 4d6 drop the lowest quickly became an option because it lessened the possible variance.

I tried having my players use the 3d6 method for stat generation in the campaign I'm starting, in an attempt to keep their scores within the bounds of what they'd get using point-buy, but the results were less than satisfactory.  Even allowing one of them to reroll three times still resulted in abysmal scores.  I wasn't happy with the results of letting them roll 4d6, either, as the results were too high, but at least they weren't as far off from point-buy.



Actually, there were four official systems for generating stats in AD&D (1e) (five, including rules from Unearthed Arcana) and none of them was straight 3d6 in order. Oddly enough, when the PHB was released in 1978, there were no published rules for rolling ability scores for AD&D. So, many people continued as they had done before (i.e. 3d6 in order). However, Gygax himself noted the importance in AD&D of having relatively higher scores (since, in the original rules, there were very few bonuses/minuses attached to ability scores), and this is reflected in the methods presented in the DMG, released in 1979.

They are:
(1) 4d6, drop lowest, six times, arrange to taste.
(2) 3d6, twelve times, choose six highest, arrange to taste.
(3) 3d6, in order, for each ability score 6 times, choose the highest for each [NB, requires 36 rolls!]
(4) 3d6, in order, to produce 12 sets, choose desired set [NB, this requires 72 rolls!]

Unearthed Arcana allowed for a fifth method, only for humans, for choosing class first then then consulting a chart. In the chart, a number of d6 would be listed for each stat (from 9d6 in the most favorable to 3d6 in least). Roll the appropriate number of dice for the chosen class and add the three highest dice for each stat.
IIRC, 1st edition, and maybe BECMI, used 3d6, not 4d6, but 4d6 drop the lowest quickly became an option because it lessened the possible variance.

I tried having my players use the 3d6 method for stat generation in the campaign I'm starting, in an attempt to keep their scores within the bounds of what they'd get using point-buy, but the results were less than satisfactory.  Even allowing one of them to reroll three times still resulted in abysmal scores.  I wasn't happy with the results of letting them roll 4d6, either, as the results were too high, but at least they weren't as far off from point-buy.



Actually, there were four official systems for generating stats in AD&D (1e) (five, including rules from Unearthed Arcana) and none of them was straight 3d6 in order. Oddly enough, when the PHB was released in 1978, there were no published rules for rolling ability scores for AD&D. So, many people continued as they had done before (i.e. 3d6 in order). However, Gygax himself noted the importance in AD&D of having relatively higher scores (since, in the original rules, there were very few bonuses/minuses attached to ability scores), and this is reflected in the methods presented in the DMG, released in 1979.

They are:
(1) 4d6, drop lowest, six times, arrange to taste.
(2) 3d6, twelve times, choose six highest, arrange to taste.
(3) 3d6, in order, for each ability score 6 times, choose the highest for each [NB, requires 36 rolls!]
(4) 3d6, in order, to produce 12 sets, choose desired set [NB, this requires 72 rolls!]

Unearthed Arcana allowed for a fifth method, only for humans, for choosing class first then then consulting a chart. In the chart, a number of d6 would be listed for each stat (from 9d6 in the most favorable to 3d6 in least). Roll the appropriate number of dice for the chosen class and add the three highest dice for each stat.



So, where did they get the "3d6 in order" method they used before?  That's what we used the first few times I played, in 1st edition.
O/Basic D&D.


I can't attest to original D&D as I have never owned it, but as for the rest of pre-WotC D&D, some things to note about ability score generation:

AD&D 1e, as mentioned above, was actually much less brutal than people blame it for being - the only place where it mentions rolling scores with 3d6 it also includes rolling multiple times because they wanted you to get higher scores.

Basic D&D encouraged 3d6 rolled in order... but even that wasn't as brutal as it sounds because you were, in the versions I read, allowed to swap a pair of scores and to reduce some scores your class didn't really utilize to raise your Prime Requisite on a 2 for 1 basis, which combined allows for most sets of rolled scores to be useful for any of that games classes you wish to play.

AD&D 2e, however, puts things to a default of exactly as brutal as AD&D is being widely remembered as being - Method I (the only method detailed prior to the heading "Alternative Dice-Rolling Methods") is 3d6 in order.

To contrast: AD&D 1e says, to paraphrase, "you could roll 3d6 in order, but you need two scores of 15+ to have a good chance to survive so you should probably use one of these other methods" and AD&D says, to quote precisely "Only a few characters will have high scores (15 and above), so you should treasure these characters."

A full flip from "you need two 15s" to "15s are rare, so be happy if you get one" - and with basically no change in what a particular score rating means for your character.

ATTENTION:  If while reading my post you find yourself thinking "Either this guy is being sarcastic, or he is an idiot," do please assume that I am an idiot. It makes reading your replies more entertaining. If, however, you find yourself hoping that I am not being even remotely serious then you are very likely correct as I find irreverence and being ridiculous to be relaxing.

So, where did they get the "3d6 in order" method they used before?  That's what we used the first few times I played, in 1st edition.



Rolling 3d6 in order comes from the way the original game was played (cf. Men and Magic, p. 10):

"Prior to the character selection by players it is necessary for the referee to roll three six-sided dice in order to rate each as to various abilities, and thus aid them in selecting a role. Categories of ability are: Strength, Intelligence, Wisdom, Constitution, Dexterity, and Charisma."

Note, curiously, that the oldest expression of the rules as the referee (i.e. the DM) roll for scores, but in practice I know players generally did. Also, while it does not explicitly state "in order" here, we can tell from the example given on the same page that they were rolled in order, and not to taste:

"A sample of the record of a character appears like this:

Name: Xylarthen Class: Magic-User
Strength: 6 Intelligence: 11 Wisdom: 13
Constitution: 12 Dexterity: 9 Charisma: 8
Gold Pieces  Experience
70              Nil

This supposed player would have progressed faster as a Cleric, but because of a personal preference for magic opted for that class. With a strength of only 6 there was no real chance for him to become a fighter. His constitutional score indicates good health and the ability to take punishment of most forms. A dexterity of 9 (low average) means that he will not be particularly fast nor accurate. He is below average in charisma, but not hopelessly so." (ibid.) [NB, a character with a prime requisite of 13-14 gained a 5% bonus to earned experience, 15+ a 10% bonus, 7-8 a -10%, and 6 or less -20%, so he "would have progressed faster as a Cleric" not only because clerics have a generally more favorable rate of gaining levels anyway, but taking into account having a 13 in Wisdom.]

So, people who played D&D back in the day just played as they had with the original sets (plus the supplements, articles from Strategic Review and The Dragon, etc.), and when the Monster Manual and Players Handbook came out in 1977 and 1978, they just used them as new bits and pieces to the game. AD&D wasn't a "complete" system until the DMG came out in 1979, and even then there is not perfect consistency across the 3 AD&D orginal rule books. People didn't mind so much if they already knew how to play, so I have no doubt that many players still rolled 3d6 in order after the DMG came out.

Also, note that, in the original rulebooks (i.e. not counting even the Greyhawk supplement), ability scores did very little, mechanically. Prime requisites helped (or hindered) experience (see my post above), a Constitution of 15+ added one point per hit die, 6 or less subtracted one, and scores between 7-12 had only a percentage "chance of surviving" (i.e. from petrification, polymorph, etc.), Dexterity 12+ added one to missile fire to hit, 9 or less subtracted one. That's it!

So, apart from considerations of the rate of gaining experience low or high scores didn't mean a great deal. Greyhawk added some incentive for high scores (e.g. high and "exceptional" Strength for fighters' bonus to hit and damage, Dexterity to fighters' chance to evade being hit, rules for chances to "know" a spell for Intelligence for magic users), but still, for the most part, a middling score was just fine, even in a prime requisite.

While the logic of wanting high ability scores to be mechanically meaningful is an old one in the game, it has also haunted the game as well, and there is good logic in making ability scores mean less, while making class and level mean more. This makes actual play more important than the sub-game of character creation (and with it, system mastery). I don't see the game going back to that simplicity, but it's worth noting.