There's lots of ways to roll up D&D characters including not rolling at all. Methods for it have changed over the years as the game itself has changed and even the values being generated have changed in their importance. With the advent of 3E the focus of the rules shifted away from ability scores directly to the bonuses from those scores. There was already a shrinking need to use them directly in prior editions but there came to be very little that the ability score in and of itself was used for in the game. The bonuses/penalties derived from it are used instead. Someday, a new version could even drop ability scores as unnecessary, though it could also revert somewhat and try to make them more significant, more frequently and directly used. Given their traditional place in the game though, they might just stick around as they are. What IS likely to change are the many ways in which those scores can be first generated. That constant change certainly hasn't slowed down just because of other changes in the game. This article was written when 3E was most popular take on the rules but the historical perspectives and methods given can still be useful for those looking at alternative methods of generating scores.
Due to how 3rd Edition is designed there was no practical difference between even-numbered scores and the next higher number. So, 16 and 17 for example are functionally identical, both giving a bonus of 3. Previous versions were suffering from "stat inflation". That is, in order to get bonuses out of stats AT ALL your scores had to be quite high, generally 15 or better. As the typical approach to long-term gameplay changed with editions it became more desirable to come up with a method of character generation that gave players more higher scores. In AD&D (1E/2E) you particularly didn't want to wind up with lots of scores within the 7-point spread between 8 and 14 that would be functionally meaningless in play (having no effect). In 3E any score above 11 will get you a bonus instead of needing scores of at least 15, and of course the lower end of the scale has been brought upward so that less than 10 is a penalty. What this means is that a better range of scores provide benefits so less extreme emphasis needs to be placed on getting genuinely high scores. However, since the threshold for penalties has similarly been brought up even as the threshold for bonuses has been brought down it has become even more desirable to have scores that are never below "average". At the same time that bonuses were made a greater focus in the game the likelihood that characters would have penalties was increased.
These changes almost can't be emphasized enough when comparing how people were trying to generate ability scores before and after. Using a 3d6 curve of probability the chances of a bonus for a given score in AD&D was only about 10%. You would be lucky to get a score that provided a bonus at all, much less even one truly high score and truly good bonus. Using best 3 out of 4d6 was better as you could be reasonably confident of at least one score with a bonus, but still not a really ideal and consistent spread barring somewhat unusual dice rolls. In 3E, using 3/4d6 as the default method your chances of a given score having a bonus exceeds 50% but the chances of a penalty are also nearly equal. Though it is often overlooked, 3E also has a minimum re-roll threshold of having bonuses totaling at least +1 and at least one stat over 13.
In-depth analysis of the long-term impact of the methods used for stat generation has yet to be done insofar as I know. Anecdotal declarations about which method - ANY method - is best need to be taken with a grain of salt. What's most preferable to you is unlikely to be preferable to everyone because we all have varied tastes and certainly varied priorities. Some will value equal playing fields among generated characters; some prefer random results; some want to accommodate the players every whim in character design; some want to challenge the players creativity with limitations instead; some DM's don't mind high scores in their game whereas others despise and fear them (rightly or wrongly); and some actually need to deal with cheating players.
As a player I personally enjoy dice rolling for generating characters. It's FUN. That's probably because it is gambling in a sense. When you "win" with several high rolls it's enjoyable. But often times to get to those rolls you have to stomach a lengthy series of really average and unexciting results and that is an inherent downside to that approach. There just isn't much in the way of die-rolling methods that can give both consistency of results and the spread of results that is ultimately desired (at least by me) while still being suitably random. Yet non-random methods are (to me) sterile, uninteresting, and actually feel to me as if they promote a certain stagnation of creativity and certainly narrow the concept of what is "fair" and equitable.
While there's more to be said regarding the game design/theory behind methods that's not really what I'm intending to do here. I'll start by listing the official character creation methods given throughout the history of the game and then add some of the unofficial ones I've seen, including one or two of my own. The methods are in color and my commentary in black text.
Any commentary regarding the usefulness of various methods is hopelessly biased and largely anecdotal on my part anyway. I'm not a statistician and to repeat - what I want from the game and character generation is NOT the same as what everyone else wants or needs. I've eliminated many of my initial comments which I, myself, ultimately deemed to be of questionable value given my outspoken biases and lack of statistical expertise. Still I just can't resist taking the opportunity to push my own agenda and views at least a little. The usefulness of this overview in any case lies in giving people a VERY broad spectrum of creative methods of character generation both old and new, not in my own opinions. When looking at various methods from a historical perspective they MUST be taken in the context of the rules system in which they were first proposed and I've tried to make that context clear. The impact of various methods in 1E have VERY different implications in 3E or 4E because the same ability scores have seriously differing bonus values - and the bonus values in turn have different impact in the rule sets they appear in. Methods that work fine in AD&D may be unnecessarily excessive in 3E/4E.
Basic D&D (Holmes):
So, while the ability generation method was quite draconian when looked at in modern terms, in the context of the rules THEN it mattered very little what your scores were because there were so few bonuses to even be gained. A very low strength fighter routinely was little different from a high strength fighter.
I've never read the later publications of the Basic/Expert/Companion/Master/Immortals rule set much less played them, but as I understand it the character creation method did not actually change much (except possibly the bonuses being granted.) It was 3d6, in order, and then modify a bit to try to salvage something if you didn't wish to play the class suggested by your strongest scores.
AD&D (1st Edition)
Best 3 of 4d6, arrange as desired.
This has been the default method that has been used in the vast majority of games I've played in or run for many years. It's probably the one that is simplest, fastest and produces the best results overall, even across multiple versions of the game. While you still get plenty of very blasé characters with this method, a little patience from the player, some willingness on the part of the DM to allow re-rolls for a more palatable character and things generally work out.
However here again we need to put the method into historical context. AD&D introduced the idea of having to qualify for a given class with minimum ability scores. Even though some other new character generation methods were allowing complete freedom to arrange the scores after they were rolled, the scores needed to qualify were sometimes ludicrously high and not even in the area where the class really needed a high score given their role. Failure to have high scores in key areas resulted in characters that could be HIGHLY ineffective compared to other characters of the same class. If you were lucky and got those high scores you'd absolutely blow out other characters in comparison to your own. For example a paladin required a 17 in charisma to qualify, but if that's your only high score you're left with a paladin with severely "handicapped" fighting ability compared to a fighter who would have been free to place a 17 in strength where it belonged for a character whose forte is still combat. By associating all the bonuses with very high stats, having those high stats became more valuable than one or more levels of a class!
This was doubly true for fighters because AD&D additionally introduced the concept of super-strength. If a fighter character rolled an 18 for strength he would automatically make an additional percentile roll for super-strength (thus, it was also known as percentile strength for those unfamiliar with 1E). It was an expansion to the "normal" chart of strength bonuses. The bonuses to hit and damage with super strength were just WILDLY out of scale with other ability scores and with "normal" strength and it really did seem unfair to see one fighter with percentile strength and another without it in the same party. Unless one dealt with melee and the other concentrated on missile fire abilities, the lower strength fighter was so hopelessly outclassed it was painful to see and easily created an understandable degree of jealousy.
As if that weren't bad enough, AD&D also introduced racial minimum and maximum ability scores, theoretically obstructing a player from playing not only the class, but the race he wanted as well.
Finally, AD&D also introduced maximum levels for characters of a given race/class combination. You could be a dwarven fighter, for example, but you weren't allowed to be any higher than 9th level - and that only with an 18 strength. NPC's were given higher limits in many cases.
Needless to say this restrictive combination of circumstances inspired a great many tweaks and experiments with character generation methods trying to eliminate or at least get around the obstacles. Now my own personal experience was that most of these AD&D restrictions were just flat out ignored by just about everybody, or at least they shamelessly sought ways to circumvent them or reduce their impact with more lenient house rules. It was the perspective of the Core Rules (a perspective supported strongly, but stupidly over the years by official proclamations) that this all maintained some kind of fair and level playing field; that the hard-to-qualify-for characters were more powerful, and thus should be more rare. The perspective was that with fairly frequent character deaths and generating new ones and with campaigns typically lasting several years balance was maintained by making such characters infrequent or even rarely seen, and players were encouraged and even expected to be satisfied with playing the characters they could qualify for.
As I said it is because of all those restrictions and obstacles in 1E AD&D that generation methods begin to really proliferate and planted seeds of changing notions of how the game could and should be approached. There was still a certain amount of wanting to find a middle ground between not violating "rules" and actually being able to play the game the way they wanted to play it. Officially there was no bending of the old perspective on the rules, no admission that they even might be overly restrictive, if not downright asinine. It would persist even all the way through 2nd Edition until 2000 and the release of 3E.
Roll 3d6 twelve times, take the best six results, arrange as desired.
A 3d6 roll in and of itself has a poor percentage chance of rolling scores high enough to actually qualify for bonuses in AD&D (about 10% chance of 15 or better) but an even better chance of a penalty (about 17% chance of 7 or lower). Rolling twelve times instead of just 6 helps to eliminate that rather notable drag.
Scores are rolled and recorded in order. 3d6 rolled six times for each ability and the highest score for each ability is kept.
Produces generally better results for high scores, and overcomes the severe likelihood of a penalty from a given 3d6 roll, but still quite likely that the character will be less than well-suited, if not simply unqualified for the class that the player desires to play. Since it requires thirty-six rolls, only a sixth of which would be used it's hardly an efficient method. It feels like you'll never stop rolling and more so if you wind up RE-rolling for whatever reason. A characters lowest score could still be Dexterity, and if the player had wanted to play a thief (or the party desperately needed a thief) the ability score generation method is getting in the way. This is a general problem with "record in order" methods - you end up rolling lots of sets of ability scores until you finally get a set that will make for a reasonably appropriate character of the class you want.
Scores are rolled and recorded in order using 3d6. Twelve such ability score sets are generated and the player chooses the one he likes best.
In choosing from 12 different sets the player is almost certain to find a set of good scores - the problem is that they are still unlikely to be positioned in a way to make a good candidate for the class he wants. Worse than method IV it requires 72 rolls only six of which will be used, if even that, because it can still too easily fail to produce a proper set of scores for a specific class. It almost seems as if a viable alternative is to simply generate characters until a set of scores is obtained that the player likes - regardless of how long it takes.
Used only for human characters (supposedly), the player chooses his class and then consults a chart telling how many d6 to roll for any given ability score. The most was 9d6 (out of which you'd take the best 3), and ranged down one die at a time (8d6, 7d6) to a low of 3d6. If you're paying attention that's 7 ability scores not 6. This is because UA also introduced the Comeliness score that isolated sheer physical appearance from Charisma and this method assumed it was being used. Furthermore, if the character did not then qualify for the class he desired his scores would be raised to the minimum needed to qualify.
This method seemed to be a response to the clear understanding that a great many players saw qualifying scores for what they really were - an unwelcome annoyance, not a true balancing factor. It guaranteed that one way or another a character could be the class that the player wanted (though of course you had to be Human, meaning you still couldn't pick both class and race in all cases and be "officially" guaranteed your combination would work). Its primary drawback is that it produced a fair number of characters that are a little TOO good (9d6 indeed!) and the characters it creates are clone characters - every human thief looks much like every other human thief because the chart is so hopelessly weighted toward not just high scores, but maximum scores in certain areas. It really does create super-characters, although I have to relate an anecdote on that score.
When we first used this system there was one player who misunderstood how it was supposed to be used. He didn't take the best 3 dice of his rolls - he added them all together. Yet his character was STILL a hopeless loser with several scores in single digits and his only good scores, while barely above average, were in useless categories for the class he'd chosen. It only goes to show that if you're not using one of the guaranteed equity systems you have to have acceptable thresholds AND still be willing to allow character re-rolls in those circumstances when one character, though meeting minimums, is highly outclassed by other PC's because of lower stats.
AD&D (2nd Edition)
Roll 3d6, in order, once for each ability score - no adjustment (except for race), no rearrangement.
This is the most insane and hardcore method ever devised. I call it the "Iron Man" method. It virtually assures little or no bonuses and statistically in 2E will most frequently create characters with more penalties than bonuses! Given the requirement in AD&D of actually qualifying for a class by needing high stats over and above what might otherwise be useful for a character it is an excessively restrictive method. It's simple and it's fast, but it's also well and truly Draconian (in the original definition of the word) as there is no explanation for why high character stats - stats that get you bonuses - should need to be so rare. Only exceptional luck allows you to get the character you may want with reasonable scores. Most of the time you have few choices about the class the character becomes or else you do a LOT of re-rolling. Classes with high qualifying scores like Rangers and Paladins are so unlikely as to not be worth really even considering as a possibility - they would be strictly freak occurrences (and this assumes that the player in question wants and chooses such a class when he CAN qualify for it.)
It should be pointed out that even in the 2nd Edition DM's Guide they were noting the severe drawbacks to various generation methods. Scores which were high enough to get bonuses (15+) were still considered to be "exceptional". That is, from an official position, most characters weren't supposed to be able to get them! The DMG text seemed to suggest that even seeking the best arrangement of scores possible and the highest bonuses possible was to be DISCOURAGED. The notion (at least the official notion) was that players were supposed to be satisfied with bad scores and less desirable characters, or that imagination should entirely supplant logical and natural desires for always seeking to make the most out of everything you got, or that it was still unfair for players to actually be allowed to play the character class and race they wanted to rather than just what they could manage with what they got. Players needed to always accept what they got in order to keep things fair.
In practice, this just never happened anymore and EVERYONE knew it. Yet it was still always the official perspective of the rules and nothing would budge the repeated assertion by anyone "official" that it was, in fact, shockingly naive and annoying. But by the latter days of 2E it was at least no longer being stressed as it once had been. It was understood and accepted at least semi-officially that the stat generation limitations were being wildly circumvented, but if push came to shove they were still presented as holy writ up until the 3E rules were released.
Roll 3d6 twice for each ability in order, keeping the better of the two rolls without further rearrangement.
Like most 2E methods it fails to address the probable lack of eligibility for a desired class.
Six rolls of 3d6 each arranged as desired.
This method would seem to at least tacitly acknowledge that players should be allowed to try to play the character class that they want, not just be satisfied with what they can get. But it's still just 3d6 and thus still keeps high scores very infrequent making them abnormally powerful when they occur. Very flawed characters are frequent results, truly good characters simply not frequent enough, and the rest are just tediously average.
Roll 3d6 twelve times, take the best 6 and arrange as desired. Identical to Method II from 1st Edition.
Identical to Method I from 1st Edition - best 3 of 4d6 rolled six times and arranged as desired.
Each ability starts at a minimum of 8. Roll 7 dice and add to stats as desired. No score can be higher than 18 and the full amount of any die must be able to be added. I.e., if you have two 6's you can't add both to the base of 8 and drop the extra 2 points in order to get an 18. An 18 is only possible by some combination of dice adding up to 10.
This is actually a rather decent method. It gives the player a great deal of control over the final character, perhaps more than Method V (or I from 1st Edition) and I think even made it probable that a character could qualify for classes like Paladin without being totally gimped. It eliminates all chance of really low rolls by having a minimum above that which provided a penalty. In fact the only way you can manage a penalty is by applying a negative adjustment for race to a minimum score. There is seldom going to be a combination that won't allow a 17 or 18 on at least one score.
HOWEVER, having more than one 17 or 18 seems to result in the great likelihood of other scores being drastically pulled down to the 8-10 range. Also, without freedom to re-roll it will most often result in highly average characters with scores no better than 13 or 14 (which, again in AD&D means no adjustments at all) if there is no attempt for a really high score. Furthermore, like all "arrange/build to suit" methods it invariably results in dull, min/maxed clone characters. Certain stats become "dump" stats because they are all but irrelevant to the predominant functions of the class being played. Unless you are a VERY committed roleplayer the temptation to min/max your rolls is overwhelming, and it's a noteworthy problem.
AD&D (2nd Edition) Players Option: Skills & Powers
A total of 75 points may be divided among the 6 ability scores, none lower than 3 or higher than 18.
Although this sort of method is the epitome of fairness it's also the epitome of vanilla. Without any random factors at all every swordsman typically had stats that look just like every other swordsman, every wizard like every other wizard. It makes the character generation process fair but sterile, and for that reason alone I don't care for it (just as I don't much care for the effectively identical Point Buy method from 3E). It suffers from the min/max temptations and just is not as much fun as rolling at least some amount of dice.
24d6 - unrolled - are distributed among ability scores. Each must have at least 3d6 but no more than 6d6. The dice are then rolled with the best 3 dice out of rolls of 4d6, 5d6, and 6d6 being used. The results are recorded in order assigned.
The problem with this method is that while it seems a little more likely to get better stats for appropriate abilities it's still random to an impractical degree. The allocation of more or fewer dice has too little effect on the outcome to make it a worthwhile exercise. (See my anecdote about Method V from Unearthed Arcana where even 9d6 was occasionally producing single-digit results. Rare and astonishing but clearly still an occasional issue.) You could again be stuck with poorly arranged, unwieldy characters like fighters with a 10 strength and 18 wisdom despite weighting Str with 6d6 and Wis with 3d6. And yet again you have issues with qualifying for the race or class you want.
2d6 are rolled and the result compared to a chart which tells you how many points you can spend on your ability scores and the maximum that any one score can be. The chart is weighted somewhat so that average results allow 74 points but perhaps limited to a 17 as your highest score; a 2 would allow only 68 points but a high of 18, and a 12 would allow 80 points but a maximum of only 15 for any one score.
This is no better than Method VII and in some ways worse as it can once AGAIN prevent you from playing the class you want to play if you're limited to scores of 15 but need a 17 to qualify for a class. Essentially it is attempting to force players to make varied characters but openly unconcerned with what kind of scores they might need for the kind of character they actually want to play. Plus it takes the only real advantage of Method VII (absolute equality of potential for ability scores) and tosses it out. So with the additional complications this method has nothing to recommend it.
75 character points are divided as desired among ability scores but exceptional strength scores must be purchased at a rate of 10% per character point.
Combines Method VII with a desire to keep percentile strength nominally in check. However it has no other advantage and the same disadvantages of Method VII. It also highlights the disproportional effect of exceptional strength in ALL methods in AD&D.
Note Regarding All the Above Methods
Some of the drawbacks of these methods disappear if they are used in 3E D&D where there are no racial minimums/maximums, no need to qualify for the minimum class scores, and where a 15 is not considered "exceptional". Some might even be considered quite viable possibilities and those in search of good methods should not overlook them based on my assessment of their failings under the rules systems where they were originally proposed.
Third Edition D&D
Six rolls using best 3 of 4d6, arranged as desired.
Same as Method I from 1st Edition and Method V from 2nd Edition. However, the thing you have to have drilled into your skill is that 3E stats are much more valuable than in previous editions. Anything above 11 is a bonus. For an old D&D grognard like me it's still hard to break that appetite for a range of higher scores. But, this method certainly provides a good range of useful scores better that it would in earlier editions. 3rd Edition also specifies that re-rolls be allowed if the characters combined bonuses add up to 0 or less, or if the character has no score higher than 13.
All ability scores begin at a base of 8. 25 points are given to purchase higher scores from a chart. The cost is point for point up to a score of 14 (6 points above 8 so a 6 point cost) and from there slides upwards to a score of 18 having a cost of 16 points.
It is reminiscent of 2nd Edition Method VII except that the higher scores are assigned an escalating value. It has the effect of making a high score more of a rarity because of the extreme cost needed to obtain it while maintaining equity between individual players. They all have the same potential as regards their ability scores, but higher scores are not as casually obtained as in Method VII. It makes having more than one high score a clear impracticality due to the extreme cost paid of significantly lowered other abilities. Two 18's or even an 18 and 17 combination is impossible. Two 16's results in remaining scores like a 13 and three 8's, or two 10's and three 9's. It seems to assume that high scores are in and of themselves so inherently problematic that only one should be allowed and even that should be at a savage cost. Furthermore it still suffers from the old drawback that all "arrange to taste" methods have - carbon copy characters. Every character of a given class ends up looking alike emphasizing the same stats (those most used in the game) and using the same dump stats.
As Standard point buy except that the player is given more or fewer points to purchase scores.
The suggested variants range from as stupidly low as 15 points - making even a SINGLE 18 an impossibility and scores above 14 impractical - to 32 points which still only slightly reduces the harshness of the restriction on high stats. In my opinion Method VII from 2nd Edition would be markedly superior to any point buy variation simply by virtue of not being afraid to let players have higher scores.
Scores of 15, 14, 13, 12, 10, and 8 are arranged as desired.
While such a character is quite clearly acceptable in power, it seems to me that since a mere 16 isn't even considered it is assumed that scores above 15 are somehow BAD and that irks me. It is this range of scores that is the default set of stats for use by DM's to create "elite" NPC's and that is what it is really geared primarily for. It only seems appropriate to even consider it if the DM is proceeding on the assumption that the PC's should never be better than NPC's at something. It also becomes an even more sterile, fun-free method than Method VII because there isn't even any math involved.
If I were to use an array method - and by preference I wouldn't - it would probably be a choice between various arrays. For example, one might be 18, 16, 14, 12, 10, 8; another would range from 17 to 10 (trading having no penalty scores for not having an 18); and one would probably give two 18's (or 18, 17...) in exchange for a more significant penalty score like a 7. Barring a campaign which specifically posits player characters as being highly ordinary or perhaps a generally low-power campaign (or at least what I would consider low power), there simply is nothing inherently BAD or unbalancing about high ability scores that warrants discouraging them much less forbidding them outright with an inhibiting character generation method like this. But then there ARE those occasional campaigns where the DM wants low-power PC's in a low-magic, gritty-realism sort of setting and this would then fit that perspective.
As the Default Method except once and only once during rolling the character can re-roll a single die out of one set of 4d6 that is rolled.
It would slightly improve the chances of at getting at least one genuinely high roll or eliminating at least one really low roll. A little complicated as it requires some actual gambling strategery on the part of the player - to know when to re-roll a low result and why. For example, with rolls of 1 2 6 6 a re-roll might be a good idea to seek something like a 17 or 18. Rolls of 1 3 6 6 however give an already good result of 15. But then what if one of the first rolls the player makes is 1 2 3 1 ? Should he re-roll to eliminate a really low result? The real problem however is that it ultimately it doesn't matter. The impact of that single die is generally going to be irrelevant to the larger game and just as likely irrelevant to the stat generation.
Roll six times recording in order using the best 3 of 4d6. Re-roll any one ability score taking the best of the two. Then switch any two scores.
In a strange way I like this. It is the method I have now settled on for use for any Old School-style campaign I'll run - though I will still continue to look for another method and ultimately bow to the players desires if they hate it. As I see it, it would likely allow the elimination of at least one low score, possibly even replacing it with a very high one, or perhaps merely improving a mediocre score a bit. It's the score switch that I like, enabling some customization of the results so that a character need not be saddled with a low score where it's most needed, but then challenging the player with a character who is almost always going to have notable imperfections, yet not generally "gimping" the character or making it unplayable. Such characters will be qualified for their job, but often not ideally qualified.
Since 3E eliminated the class and race qualifications this method now enables what Method I in second edition (the "Iron Man" method) must have been intended to do - challenge the players with characters that need to be taken as-is, warts and all, and exercise the imagination to account for and overcome their inadequacies. Even if players came to me and floored me by asking for a really hardcore campaign this is almost certainly the method I'd use rather than Iron Man. But given a reasonable DM who allows a few re-rolls of overly-average characters it will produce fine results and seems to me to be the best compromise of all possible stat-generation methods - for Old School approaches to the game.
There once was a time where the 1E/2E perspective about maintaining balance had me convinced that players shouldn't have complete control over their characters creation in ways that I now consider almost fascist. Though I never inflicted it upon any players I secretly held the belief that the only proper way to roll up characters was the Iron Man method (3d6, in order, six times). Only after 3E had been out for a while did I really embrace the notion that, as somebody once put it and convinced me otherwise, "It is not a crime for PC's to be good at something," and start to rethink the very fundamental reasoning behind character generation methods.
Six rolls of 3d6 arranged as desired. Also, the character may be re-rolled only if the bonuses add up to -3 or less or if there is no score of 12 or higher.
The allowance to arrange as desired is good given the low values produced by the harsh 3d6 bell curve, but the lowered threshold for re-rolls is whacked. Though it acknowledges that the average rolls are going to be much lower, a total of -2 of accumulated bonuses (which would NOT merit a re-roll) is a horrible array of ability scores and indicates a character who, despite the name of this method, is BELOW average in his capabilities. While very hardcore roleplayers, realists, sadists and masochists may like this method I think it blows - for ANY edition - and I can't fathom what the hell they were thinking when they included it.
3d6 rolled six times and recorded in order. The character may be re-rolled if the bonuses add up to -3 or less or if there is no score of 12 or higher.
Only marginally less hardcore than Method I from 2E but only because it provides a threshold for re-rolling. Compared to the Customized Average method just above it's worse in that you'll need to either play really low-grade characters of classes you didn't really want, or do a lot of re-rolling to get something palatable - even by THESE low standards.
Six rolls using the best 3 of 5d6 and arranged as desired. Re-roll if modifiers total less than +2 or there is no score of 15 or higher.
I'd go for this one if it weren't for the fact that it really does make high-powered characters for 3E. Low scores are TOO rare for my tastes and high scores become too common. What I'd prefer if such could be found is a method that generates a spread of marginally lower and generally higher scores on a consistent basis while still being random rolls of dice. It's still no guarantee of Supermen but given the easy access to bonuses in 3E rules it's close. Possibly useful for a really tough game though I wonder if scaling up the opposition isn't a better approach to a tough game world than Uber-ing the PC's.
Methods from Other Sources
DM simply lets players choose their ability scores out of the blue. No rolling, and generally no maximums or limitations except the players own good sense about what is reasonable.
Not at all a common approach but IF you can trust your players I don't see much problem with it over other methods.
Players make 7 rolls using the best 3 of 4d6, then drop one of the 7 and arrange as desired.
Not a huge advantage over the Default method but for some it might provide that slight improvement that they want to see.
Each player uses the same set of 18 individual die results, adding as many of them to an ability as desired so long as no leftover points are wasted (similar to 3E Method VII) and results for all scores are between 3 and 18. A typical pool would be 6 6 6 6 5 5 5 5 4 4 4 3 3 3 2 2 1 1, though it could be weighted as the DM sees fit.
This is a method that I think is palatable (if unexciting) for a "diceless" method. Assuming a pool as indicated it allows more than one high score to be engineered but without driving remaining scores unduly far into negatives: low or average scores are needed to obtain the highs. Its only real drawback is again the complete removal of the random element making the exercise a sterile and calculating one. A slight modification would be to only allow 3 "dice" to be combined for any ability score.
A logical modification of the Fixed Dice Pool where the 18d6 (or perhaps more or fewer dice) are rolled all at once and the results used by all players to combine into 6 ability scores of between 3 and 18.
Better than Fixed Pool as it reintroduces the random element. Again, a modification would be that only 3 dice can be combined together to create a given score. Also again it retains a lot of the sterility and tendency to unoriginality in character design of free-arrangement methods. Yet another variation has the players each rolling one stat, perhaps making multiple rolls until 6 scores are arrived at which are then used by all players for creating their characters.
This method is commonly executed with playing cards though it is essentially a Dice Pool method. The way it works is that cards from a standard poker deck are used to represent dice rolls. Ace=1, deuce=2, etc. In one option there are 4 suits from ace to 6. They are shuffled and six separate "hands" of 4 cards each are dealt. The best three cards are kept in each hand and the total value of the hands equals the ability scores and they are then assigned as desired. Other options generally just alter the available pool values (and may require more than one deck of cards). For example, a pool of 1 ace, 2 deuces, 3 threes... on up to 6 sixes and then a joker or two added in. Hands of 3 cards are then dealt whose totals are the ability score values, with the joker being a "wildcard" that doubles the value of any one other card that is dealt in a given hand, or provides some other bonus effect.
Another example has a single suit of cards from A-K. The King is set aside and six sets of 2 cards each are dealt for the six ability scores. Player then assigns King as desired. Card values are A=1, 2=2... 10=10, J or Q=8, K=5.
Not really a method on its own as such. It is just the combination of two or more random generation methods. One method is used to generate ability scores and then a second, often entirely different method. The best scores or best SET of scores is kept. One such example is allowing players to use the standard method to generate scores randomly, but then if they don't like the results to opt instead for a standard Point Buy.
21d6 are rolled. The best 18 dice are kept and freely combined; 3 to each ability score. Obviously modifiable to any number of d6, not just 21.
All scores except the prime requisite for the class are rolled with 3d6 in order. For the class prime requisite (possibly more than one) you roll 3d6 but use a variant die reading method: any 1=5 and any 2=6 for example. This boosts the average of a d6 roll up to 4.8 (from 3.5), and the average of 3d6 read this way is about 14.
Charts are devised using a wide variety of dice types, number of dice, and added points so that each race has a unique outlay of dice rolls to make for generating their stats.
We didn't call it anything at the time - we just made it up and used it as a different method from what we'd always used. I don't know who made the first such charts. It had to have been either Casey N. or Scott A. from the group I played in. This was for some of the REALLY old games of AD&D that I was involved in where more variety and better stats were desired for character creation. I think it may have even migrated from OD&D to 1E before eventually being supplanted almost exclusively by the 4d6 method (with a few short forays into other methods). You'd look at the line or column for your chosen character race and find the dice combinations to roll for each specific stat.
Sadly, no original examples of these charts actually survived the era. The only specific examples I can recall are that a dwarf would roll 4d8 for strength, but only 3d4 for dexterity. Dwarves were highly favored in these games, even OVERLY favored, but their weakness was always Dex (or more accurately Agility since for the campaigns in which this saw use we frequently split Dex into two scores - Dexterity and Agility. Why? Someone thought it made sense at the time I guess...). Combined with a penalty to ability checks from wearing heavy armor, dwarves became like turtles turned on their backs if they were knocked down. There was originally no attempt to maintain equity of ANY kind between character races and score values with this method.
Though it deals with all standard races and not just humans it suffered from the same drawbacks as UA method V - primarily that would be that the weighting was still far from any kind of guarantee of appropriate stats. Furthermore, as suggested there was a great deal of personal preference in the charts. But I must admit that it has a certain twisted, Old School, we-were-too-young-to-know-any-better feel to it. Sure it was whacked from a game design standpoint but WHO CARES? We enjoyed it. Well, mostly enjoyed it. Unlike many games of that era where Elves were SuperMen, in these charts they were (to my recollection) disadvantaged. Still I'd go back to it in a heartbeat if anyone else actually wanted to do it, just for the gonzo fun factor.
Without a surviving original example it can't be analyzed objectively and as noted I can't recall clearly more than a few of the specifics. But then that wasn't the point of that method either. It was intended to promote the biases and preferences of the DM and players involved. Even if I did have a copy, we were always tweaking such things at whim. Using the chart to roll up a character didn't mean that another player wouldn't wind up using a different chart to roll up HIS character the next week because it had been changed by the DM in the meantime. I'd say make up your own if the idea intrigues you but I've actually re-invented the thing here anyway just for grins. I do sort of recall it being weighted to the ability score ranges suggested in the 1E racial min/max charts so I used that as a starting guide. I recall the original charts being significantly more munchkin than these, and with simpler combinations of dice, but I just couldn't bring myself to get as truly puerile as that original was. [For example, as noted, the Str roll for Dwarves was 4d8. That would have been capped at 18 and followed up with a percentile roll for Super Strength, but the AVERAGE of 4d8 is 18, so...] Obviously, this would NOT be used in combination with racial adjustments.
AD&D (1st Edition)
AD&D (2nd Edition)
Using best 3 of 4d6 (or strait 3d6 or whatever), generate 36 ability scores, putting them into a 6x6 matrix. The player may then select a set of 6 ability scores by reading in a straight line anywhere on that matrix, whether that be reading a row forwards or backwards, reading a column up or down, or either diagonal backwards or forwards. That gives 28 separate combinations that a player can choose from 36 generated scores.
For example, using 3 of 4d6 I got:
In the example matrix above look at the only 18 that's there. Take row E left to right. 15 Str, 18 Wis. Not bad for a fighting cleric. Take the same row right to left. 18 Dex, 15 Cha. Good for a lying rogue. Or concentrating more on the Cha with the Dex as a bit of a fluke you can think in terms of a missile-using paladin, or a defensible sorcerer or bard.
Now look at column E reading top to bottom. You have two great scores in Wis and Cha but you have some noteworthy weakness everywhere else. Read it bottom to top and you have a good, stereotypical mindless fighter with 17 Str and 18 Dex, although the 9 Con might be a tough pill to swallow. Taking column D from top to bottom might be better for that because it would give you lower Str and Dex but better scores over the remaining stats.
Looking for the highest Int I can get for a wizard character it would be row A read from left to right. Reading it right to left it becomes the best Con score available.
This is an interesting method that has much to recommend it. I think one of the things I like most is the innate "handicapping" that it requires. In finding the best set of stats for your desired character you almost certainly have to make some compromises, but you have a choice among which set of compromises you prefer. In that respect it is like the Organic Method, which I still prefer, but this is a good second option and I personally would have no trouble using it.
Make three stat rolls using best 3 of 4d6. For the last three stats subtract your first sets of rolls from 25. Maximum of 18 on any stat So if you roll an 18 one of your stats will be 7 (25 minus18). 7 will be the lowest mirror score you'll get from a high initial roll. But if you roll a 7 or LOWER your corresponding stat from that will be 18, making 18's much more common.
I've seen it (or similar variations) mentioned a couple of times now and it's intriguing. I'd personally consider adjusting this due to the increased number of 18's that are certain to be generated. Maybe subtract from 23 instead of 25. 18's will still be more likely than they otherwise would except that there would also be the low-end of having stats of 5 to go along with them. I'd have to leave it up to the players to tell me if they thought it was worth it or needed adjusting, but it certainly has interesting dynamics that I think have potential.
There is the assumption here that scores would be arranged as desired after being rolled, but that could also be changed to spice it up more. For example, as part of the generation process you could have players determine two stats at a time. They would pick any two stats, roll dice, and then assign the roll and the mirror result to their choice of the two stats. Or they could pick THREE stats that they will roll dice for in order, and then apply the mirror stats for those rolls as desired among the remaining three.
Another variation on the method would be that rather than rolling dice at all the player simply PICKS his desired stats and the associated mirror results. Even if a player does pick high stats they're guaranteed to get low-grade mirror stats to offset them. When subtracting from 25 to get the mirror you have to pick stats of 15 or lower in order to NOT get stats with a penalty.
One die is replaced with 6 points. That is, roll 2d6+6 for each stat. Either you can use it with free arrangement or take it in order.
Gives a minimum stat of 8, max of 18, and an average of 13! Definitely pushes down that bell curve, but I think it makes 17's and 18's a little too frequent and low scores with actual penalties possibly not frequent enough. Probably okay for a rather higher powered campaign.
Roll 6d6. The results of the dice indicate which one of the attributes is increased to provide an additional bonus from that attribute. So, a '1' means Strength is increased to provide a +1 bonus (that is, it must be a 12 Str.), a '2' means +1 to your Dex and a corresponding increase to your actual Dex attribute, and so on. If a roll makes a score go beyond 18, re-roll that die.
It has the "advantage" of utterly eliminating the possibility of a negative score. It has the disadvantage of creating characters that are inadequate for a specific class - if you don't roll any 1's or 2's your Str and Dex are not increased and you have a piss-poor fighter if that's what character you wanted. A variation on this is for the DM to predetermine a given value for the total of bonuses such as 7, 9 or 11 depending on the desired power level. Players then choose their stats freely so long as the total of bonus values is not exceeded.
A standard 3x3 grid is made with Str, Dex, Con being rows, Int, Wis, and Cha being the three columns. 4d6 are then rolled (or the usual variations thereof) nine times to fill the grid. The player then selects 6 out of the 9 possible scores from either the appropriate row or column for the given stat. Each roll can only be used once so, for example, if you have an 18 in the center square you could use it ONLY for Dex or Wis, but not both.
This is another method I find interesting due to the handicapping dynamics of it.
The player chooses 2 high-, 2 mid- and 2 low-priority stats based on what kind of character he’s aiming for. For the high stats he rolls 5d6, the mid 4d6 and the low 3d6 each (best 3 dice each as usual). These are rolled and recorded in order with no swap or trades. Take what you get first and every time.
Weights probability in favor of what’s desired but will occasionally force player out of the Same-Character-Stats-EVERY-Time rut (which was the express intent of the person I saw proposing it.)
Player rolls 4d6 and DM rolls 4d6 secretly. Player then looks at his own roll and decides to either take that roll or go with the dealers (DM’s) result instead. Repeat once for each stat.
Like the Dealing Cards method this might work best if it is somehow in keeping with the general themes or flavor that the DM wishes to emphasize in the game.
The DM creates a variety of different arrays that need not be of similar "power" levels. Players are given an XP value and then bid on the arrays with their XP. Leftover XP are awarded to the PC when play begins.
It seems like a lot of variations should be possible with this. Bidding with XP debt instead of a fixed amount of bonus XP springs to mind. Arrays don’t need to be too balanced with each other as players will naturally decide what the stats are worth to them. Those who place exceptional value on stats will advance slower than the rest. Those who take less-than-stellar stats will have a head start on others. The amount of XP initially awarded can be increased or decreased to allow for higher bidding. I'd actually like to see this tested to see how this might be viewed by players over the long term.
All stats have a base value of 6. Players roll 6d6 and assign individual die results as desired. Players then roll an additional 1d6 once for each stat in order. For more player control switch the last two steps so that the players Assigned results are done after the Random results are known. Another variant makes it the BARB Method (BAR+Bonus): Total the ability scores. If 78 or greater: No bonus points. 72 – 77: 2 bonus points. 66 – 71: 4 bonus points. 60 – 65: 6 bonus points. 54 – 59: 8 bonus points. 48 - 54: 10 bonus points. Add bonus points as desired including exceeding 18’s.
Standard 4d6 generation is used. Players may generate as many characters as they like but each must be a fully filled-out, completely kitted and ready-to-play character sheet. Keep the one you want – the rest go to the DM. That's the Hard Work part. The Luck part is that PC’s also get a Luck stat that is reduced for every PC they create but DON’T use. I.e., taking one of your first characters gets you a better Luck score, but to keep going in search of that ultimate set of stats means your PC will get progressively crappier Luck.
The point to this method is to have the players providing the DM with some ready-to-use NPC's. The downside to this method is that for this to work it must be done AT the table at the time of character creation. In 3E character creation can take a significant amount of time for some players putting them at an unfair disadvantage. And besides, if the DM is THAT hard up for prep time and game tools there are bigger issues needing to be addressed.
Roll 1d6 sets of 6 numbers using 3d6. That is, to generate a set of 6-36 numbers. The player then chooses any continuous string of 6 numbers but must record them in the order they appear. The string of numbers is to be considered a continuous loop so for example if the player chooses to start with the last number in the set as his Strength then wrapping around to the beginning of the set he takes the first 5 numbers in order as the remainder.
This is another method where the player gets to handicap his stats but doesn't have complete freedom to arrange things - he has to find the best set of numbers for what he wants.
Other General Modifications
Some modifications, mostly to the various random generation methods, repeatedly pop up.
My Personally Preferred Methods For Character Generation
Now I have my biases regarding character generation methods and ability scores which should already be obvious. I like to see characters with at least a 15 in something if not more than one. That's probably a relic of my AD&D upbringing in the game. I must admit though that there is something strangely compelling about pure Iron Man rolling (noting that its use only makes sense in 3E or original D&D) but I wouldn't go there for just any old campaign. What high scores meant for earlier editions still seems to mean the same for me even though I know better. I also despise the idea that absolute equity among characters is somehow a requirement. I want to see some characters that truly ARE better than others and for players to at least occasionally accept less-than-perfect characters as a enjoyable challenge rather than an imposition. It's a fallacy anyway that guaranteeing identical value in stats actually makes characters equal or that it affords them equal opportunity. Different class needs, different skills and feats chosen, different magic items and equipment, player skill and enthusiasm, in-game events, and the preferences and biases that the DM may have in creating and running his campaign all combine to make ability scores a vastly more- or less-important factor than some would assert.
I like to roll dice for character generation. I think I mentioned that already somewhere above. Randomness in generating a character is just more fun than simply engineering a character to me. And also I've come to dislike the innocuous-looking steel trap that is arranging all ability scores freely - the first step in what is known as min/maxing, a.k.a. maximizing the lethality index, a.k.a. power-gaming, and of course it's practically a requirement of munchkinism. It is so highly illogical and at cross-purposes to willingly handicap a character when you don't have to by NOT placing scores in only their most optimal positions (using the least used stats as dumps). I've become convinced that at least some mandatory take-the-rolls-as-they-come is actually of greater benefit to creativity and interesting, original characters than complete freedom to min/max. YM will obviously V. Min/maxing your very core ability scores was NECESSARY in earlier versions in order to get bonuses that meant anything to the character class you were creating. That just isn't true anymore in 3E where scores of 12+ receive bonuses.
Players should be faced with some degree of challenge in having to create a character with "flaws" that are not of their own choosing. Being able to play the class and race you desire, and having your character be competent at the class chosen by having at least one appropriately placed high stat should be generally enough. Some stats, or the placement of those stats should be beyond the players control at creation if more realistic characters are ever going to be seen in a campaign rather than identically min/maxed, sterile, uninspired clones. As a DM I saw such characters all through the '90's while running 2nd Edition AD&D and I hated it. I should note that I consider it likely that many of the same people who spew vicious, unnecessary attacks on generation methods that result in a PC that has "flaws" that a player has to accept will say glowing things about introducing "Character Flaws" for players to use.
There should be a minimum acceptable level for stats for ANY method that includes actual random elements. And I'd emphasize the word "minimum" there, because more importantly the DM must be able to exercise authority to allow players to re-roll characters even if they meet "minimum" standards. Why? Because random generation is random - it isn't guaranteed that the results created by more than one player will be appropriately compatible with everyone. A character with nothing below 15 can become a monstrosity when run alongside a character who has a 16 as his best stat and nothing else above 12. Players must be informed of this and freely accepting of its implications before rolling characters begins. Absolute equity among character ability scores is nonsense, but that doesn't mean that equity is completely irrelevant. That is what the DM is there for - adjudication of situations that are not or cannot be covered in detail by the rules.
A few last notes. First, as mentioned there is still something to be said for equity in stats when there is great disparity in stats, especially between PC's of the same class. Just as super-strength characters in AD&D blew any "normal" strength fighter out of the water, a fighter in 3E with an 18 Str and nothing else below 15 can unacceptably overwhelm a fighter PC with a lesser strength in the same party who might otherwise be considered to have quite adequate stats. A player who IS lucky enough to create an Uber-stat character with random rolls can easily slide into the trap of thinking that it allows them to run the show for all the PC's and hog the spotlight from that point on. Such a character requires MORE restraint and GREATER maturity from the lucky player in order to keep the other players of lower-stat characters interested in the game and not overwhelmed and buried by an overeager super-character. Just as a DM needs to allow an uncharacteristically low or average set of rolls to be re-rolled, he also needs to either declare a super-character unfit for mixing with lesser characters or have a heartfelt chat with the player about keeping his stats and his roleplaying in perspective - and laying out some potential repercussions if it becomes a problem.
Secondly, different concepts about a game world may lend themselves much more strongly to certain methods. For example, a game where all the PC's ARE clones would scream for using a default array of some kind. A grim, gritty, "realistic" game setting would lend itself more to more stringent "Iron Man" methods, as would games where an emphasis is intended to be placed on player creativity and adaptability to run ANY kind of character depending on what they can "qualify" for. And those unfortunate DM's who actually DO have to worry about cheating players need to turn to methods of character creation that emphasize equity and verifiability.
Thirdly, unless the DM has VERY compelling reasons to limit the maximum ability score to something below 18 then ANY method used should be able to create a character with at least ONE maximum score - an 18 before making adjustments for race. If it won't even allow you what Iron Man 3d6 will potentially allow you then there had better be genuine willingness on the part of the players to accept that restriction, and good reasons from the DM for asking for such restrictions. That's my opinion and you're welcome to it. One such compelling reason to do otherwise might be that the campaign is going to be very low-power, very gritty, and "realistic". An 18 would be out of place simply because everything else is being restricted to more of a real-world human-normal range.
And the winners are...
Now as a DM I still acknowledge that even that is less than perfect, and even with the re-roll threshold there will undoubtedly be characters that just don't quite cut it. In EVERY method that includes random generation I think a DM needs to be willing to allow players to re-roll again after all is said and done. If nothing else it's an acknowledgement that one player could create a mere technically-passable character while the simple vagaries of dice have provided other players with quite notably superior stats.
As an example of the organic method, take a player who wants a paladin character. Dice are rolled and he has str15, dex14, con16, int10, wis13, cha7. The low Cha is re-rolled but unfortunately is no better. The player would ideally want both high Str and high Cha, arguably the two most important stats to a paladin. If left to his own devices the PC would have almost certainly used Int as a dump and ended up with str16, dex13, con15, int7, wis16, cha15 (or close to that). Extremely vanilla, uninspired and uninteresting. However, as-is the character has an adequate Str but unacceptably low Cha for a paladin. The player doesn't want to switch with his best current stat of Con because it would leave him wanting for hit points. So perhaps instead he accepts a switch from Dex leaving him with str15, dex7, con16, int10, wis13, cha14. Essentially, the player has a choice of taking a flaw in dex or con to bring his cha to a preferred level. It's a hard choice but the end result is still a VERY playable character that nonetheless has interesting flaws that the player would never by choice have built into his character. The character could obtain dex-enhancement items to cover up his flaw or as he goes up levels dump his stat increases into Dex for a more permanent solution.
If the same stats were rolled for a wizard character it wouldn't be a hard choice to switch Con and Int, leaving a wizard who can actually do some good if he has to with a dagger. Or he could switch Str and Int producing a wizard with a superior number of hit points than would be usual. Both FAR more interesting choices for a character than what a player would opt for if left entirely to his own devices to min/max.
Return to D&D Page
Return to Home Page