Tuesday, April 5, 2011

Ability Score Handicapping

The random generation of ability scores seems to be one of those fault lines that divides even the old school gaming community, with plenty of people willing to argue its rightness or wrongness, as I recently discovered. And it's an old and active fault line, probably going all the way back to the dawn of the hobby, when someone decided he really wanted to play a magic-user and complained bitterly to his referee when he rolled a 5 for his character's Intelligence score.

By the early 1980s, there seemed to be two "solutions" for this "problem." The first, pioneered by Superhero 2044, was a point-buy system that abandoned randomness entirely. This is the mechanical birth of character creation rather than generation. The second solution is to "handicap" rolls for ability scores in order to give PCs an edge. Take, for example, this table from FGU's Space Opera, first published in 1980:
Ability scores in Space Opera (called "personal characteristics") range from 1 to 19, but, rather than generating them by rolling 3D6 (or even 1D20) with modifiers, they're generated by rolling D100 and consulting a chart that, in the words of its designers, is "'loaded' toward the higher end of the scale." That is, even a middling roll is likely to (in most cases) give a personal characteristic score of 13 or thereabouts, ensuring that PCs are generally well above average -- and that's not even counting the dice modifiers each class gets in order to boost the characteristics most closely associated with them.

TSR's own SF RPG, Star Frontiers, opted for a similar approach to generating ability scores, although, unlike Space Opera, it not only ensures that PCs are well above average but that they are not typically too far above average:
Other TSR games also attempted to blunt the edge of random ability score generation, but did so in a different fashion, as did 1980's Top Secret:
Instead of wholly divorcing the percentile roll from the ability score generated as both Space Opera and Star Frontiers do, Top Secret instead provides a sliding scale of bonuses to the roll, so that very poor rolls are given a large boost and better rolls are given a smaller one (if any at all). This is the same approach taken in 1982's Gangbusters. In fact, it's the exact same table.
I've begun to wonder if perhaps elements of these tables couldn't be adopted in Dungeons & Dragons, at least in AD&D, where ability scores matter a great deal more than they do in OD&D. What I like about these tables is that they still preserve varying degrees of randomness rather than out and out making PCs simply better than average, as do the more ridiculous methods presented in Unearthed Arcana and even the DMG.


  1. Hmmm...


    Overall Chart


    Not at all dissimilar to what Gary did with rolling stats in Dangerous Journeys... each of the basal stats was rolled using 2d6+6, for a range of 8 to 18, with an average of 13.

  2. Whoa. Rolling a 25 in TS/GB is worse than rolling a 26? You would at least expect the handicapping function to be piecewise continuous.

    The Space Opera rules appear to be a hybrid generation/creation. Each character class gets a pool of points to apply, and gets to apply them after the die roll.

  3. This comment has been removed by the author.

  4. Not at all dissimilar to what Gary did with rolling stats in Dangerous Journeys... each of the basal stats was rolled using 2d6+6, for a range of 8 to 18, with an average of 13.

    The problem with that approach is that 2d6 also makes the extremes more likely. The fall off is linear from the mean, instead of a nice bell-shaped curve. You would do better with a "3d6+3, cap at 18" approach if all you wanted to do was to adjust the mean.

    EDITTED: error from previous version.

  5. In Unearthed Arcana, Gary and co. came up with a weighted system of rolling for stats based on the class you chose. Thus if you chose a thief, you got more dice to roll your dex than a fighter who got more dice pooled to strength. It got a little wonky once you got into classes that had more than one prime requisite however.

  6. For the Star Frontiers example, there's no distinction made between PCs and NPCs; i.e., presumably NPCs would use the same table. The table is really giving a bell-shaped distribution to ability scores, not giving a boost to PCs.

    In fact, the resulting average ability score is actually less than 50%. Median: 45, mean: 47.25.

  7. I have looked at a few weird ways to roll stats but overall part of the thing with 3d6 is that it is all there, no graphs or extra numbers to remember. You can technically generate that with a single d6. It is also why methods like 4d6 drop lowest are popular, which is because it is just about as easy to remember. If you want stats that range from 9 to 18 I find 3d4+6 to work quite well:
    I use AnyDice to help visualize various dice configurations. Here is a link to to my example: http://anydice.com/program/6f7
    If that address doesn't work just take off everything after the "com" and try again.

  8. The free RPG Crimefighters that appeared in Dragon magazine uses a table that's almost exactly the same as the Top Secret / Gangbusters one.

  9. The Top Secret/Gangbusters certainly skew results to "better than random average" with a Median of 63 and a Mean of 63.5.

    I'm not sure I really like their methods. You actually do better rolling an 18 than a 27. Doesn't seem to make intuitive sense. They could make this all much simpler if they want stats better than a random roll would give you.

    Anyway, I still prefer random rolls...just like life, but I'm considering trying out random rolls plus giving the PC each 6 points or so to put into there stats as they see fit. That way, if you really wanted to be a magic-user, but were born a bit-less-than-bright, you could at least say you've studied hard for years and got your Intelligence up to a playable 12 or something. Has a bit of "real-life" to it...never mind the magic and dragons.

  10. The only reason I can think of to even bother with any sort of handicapping system at all is just to raise the minimum scores PCs can end up with without affecting the probability of them getting very high scores. And, if that's all you want, why not just say that any roll less than some number is treated as that number instead?

    Myself, I still prefer PC rolling systems that just lower the probability of them ending up with very low scores, but don't eliminate it, while also raising the probability of them getting very high scores, but don't guarantee it. To that end, I like having players roll all 18 of the dice for all of their stats at once, then letting them arrange those dice in 6 groups of 3 however they want.

  11. Ed, that's the system I've used in the past (playing AD&D 2e. The only other systems that I've used at 4d6 drop the lowest (sometimes with a re-arrange order to suit), 3d6 (sometimes with a re-arrange). Lately I've been using the standard rules from the RC, which allow some rearrangement.

    I wouldn't want to eliminate the possibility of PCs having perhaps on characteristic below average. Remember, you've got to get down to 7 or below before we're really talking about being noticeably clumsy, weak, stupid, etc. (whatever the players think about a 9).

  12. I prefer allocation method instead randomize. But I have a method from earlier.

    You roll a 6*3D6 for example and substract each result from the maximum (currently 18) and sum up the all results, divide by 6, round it and allocate.

    For example:

    The rolling results: 4, 6, 10, 13, 16, 18.

    The subtraction results: 18-4=14, 18-6=12, 18-10=8, 18-13=5, 18-16=2, 18-18=0.

    Sum up: 14+12+8+5+2+0=41 -> round(41/3)=14

    You can allocate additional 14 points to the rolling results. But the maximum is still limit.

  13. First off...DRBargle...thats not Bargle in your Avatar thats Baron Von Hendriks, Cruel Master of Fort Doom and the Black Eagle Barony. Thats been driving me nuts.

    As to this whole variance of Stat Generation:
    I've no problem with simply having PCs roll 3d6 and alocate each roll to an individual ability.

    Thats how it was in the earliest D&D. I suppose the common whinge is when the player wants a Class and Race that they dont have the Stats for. There is nothing wrong with that wretched Bell Curve of Statistical Probabilities that comes with rolling a 3D6.

    I have no problem with Players throwing their Self Generated PCs into a pool on the Table and selecting the One they want - allowing them to negotiate with each other for rights. Hell I even allow them to throw the Total Starting Gold into the Common pool as a method of buying the PCs they want.

  14. Sorry, yes. Bargle is in the bottom left-hand corner of that picture. But I don't know why I haven't changed it yet.

  15. I don't think that any method of skewing roll results really solves the problem.

    Players seem to compare their stats to other player characters, not to NPCs or to the range of possible results. So making 13 the new average just means someone with 11 now feels like they have a bad result. It also makes a 3 even worse than it would be if 10 was average, which can lead to more attempts to skew the results, which...and so on.

    Point-buy leads to various bad results which I'm sure everyone is aware of.

    I'd just as soon players not roll or pick attributes at all. For example base STR, DEX and INT could be based on class, and WIS, CHR and CON based on race. Characters of the same class could be distinguished by having a 'background' which gave a small bonus to certain tasks that weren't the preserve of any class or race (I'm thinking of things like 'sailor' or 'gambler' - Akrasia's blog has a system for this).

    However I can see how this would be 'not D&D any more' to a lot of people.

    At the moment I'm running an OD&D game, where attributes have minimal impact.

  16. Every handicapping system mentioned here so far seems to me like a lot of trouble for little reason.

    Why not just declare that any PC stats rolled lower than some number (possibly different numbers for different stats) are that minimum number instead, and just be done with it?

    Or, if you want most PCs to have very high stats, why not just add some number to every PC stat roll (again, possibly different numbers for different stats), but treat any result greater than 18 as 18?

    Why bother with any sort of complicated table or calculation? What advantage do any of them have over either simple minimums or simple point adds?

  17. I've always liked the Metzner Basic/Rules Compendium solution to this problem (don't recall if it's in Moldvoy's Basic rules as well or not): 3d6 in order but you can shuffle points around among attributes on a weighted scale basis. Keeps the randomness but in most cases will let the guy who really wants to be the smart magic-user be the smart magic-user.

    I also like 3d6 put them in whatever order you want - that actually ended up being the default for the longest running campaign I ever ran and it worked out quite well for everyone involved. I'm also tempted to port over the Mazes and Minotaurs version of this rule the next time I run a D&D game - choose your class first, then roll your attributes. Highest scores must go in the prime requisite for your class, arrange others to taste.

  18. "I don't think that any method of skewing roll results really solves the problem [because p]layers seem to compare their stats to other player characters, not to NPCs or to the range of possible results. So making 13 the new average just means someone with 11 now feels like they have a bad result. It also makes a 3 even worse than it would be if 10 was average, which can lead to more attempts to skew the results, which...and so on."--anarchist

    If that's true -- if real problem is that players' childish egos can't bear any of their characters ever being below average in any way -- then the real solution can't be game-mechanical. It has to be social. And it'll probably require forcing players to choose between either playing characters who are in some way below average or not playing at all.

  19. To echo my post to the last discussion, replace your dice with playing cards!!! Allows for very flexible handicapping, ensures randomness, yet leaves room for a little player agency.

    It's the absolute best character generation system you've never seen printed in a book.

  20. I was inspired by the first table to do some theorycrafting and came up with my own table for generating D&D ability scores using a d100.. Available here if anyone fancies a peek: http://bit.ly/i5X9o3

    In the last 6-7 years (since I've been playing D&D of any sort again) I've only ever used dice to create a character twice and even then we used variant methods. Generally we do points buy methods and whilst I like that way of doing it I do sometimes yearn for a simple 3d6 in order.

    I was too young to appreciate the challenges associated with playing a character created that way when I did it (if I ever did..) but I think now would be a good time to give it a whirl. If I get my homebrew campaign off the ground I'd use 3d6 in order unless my d100 method takes my fancy!

  21. I absolutely agree with Ed above that this is a social problem...But just for the heck of it, here are a few more things you can try:

    Roll 2d6 and subtract the *lower* die from the *higher*, then subtract 1. You will get a range of numbers from -1 to +4, with a heavy bias towards zero and an assymetrical curve towards both extremes.
    Those are your bonuses. Then just retroengineer the stats (-1=8, 0=10, +1=12, +2=14...). If you're missing your odd results roll another die or flip a coin or something.

    Still pretty random, but again, no scores below 8, maximum is still 18, and the expected result should be ~13. Pretty powerful.

    *Give the players a big pile of d6s, at least 18 per player, but possibly a few more. They roll all of them, then spend the individual dice to add to a score. It's half-way between point-buy and random. Oh, and scores sill cap at 18, so even if you spend more pips on it, it's still going to be 18. Generally lets players to be really good at the thing they want to be good at, at the cost of other stats. Expect far more 18s as with the traditional method, but the rest of the stats will be weaker. Probably fairly unbalanced.

    *6d6 in order
    Your highest die gives you an 18, lowest gives you an 8, second highest gives you a 14, the rest give you 10s. (replace example scores with whatever "power level" is appropriate for your game)

    *something I did once: everyone rolls their 3d6 seven times (so there's one discardable result), but then they can 1) shuffle the results amongs the scores 2) trade results with other players. So if Bob got two 16s and wants to be a fighter, he can trade them away to Joe for an 18 (which he puts in Str), while Joe needs two high stats because he wants to be an Elf.

    I think all of the above are somewhere halfway between "totally random" and "customizable to taste".

  22. There's another stat-generation method that hasn't come up here that's truly bizarre:
    Villains and Vigilantes, if played by the book, had players playing THEMSELVES with superpowers as PCs. Your character's secret identity was supposed to be YOU. Stat generation was supposed to involve the player and the GM mutually agreeing upon what the PLAYER'S strength, intelligence, charisma, etc, would be.

    I really wonder if anyone ever played it that way. Me and my pals never did - we just rolled 3D6 or 4D6-lowest roll.

  23. We didn't play much V&V, but we actually did stat our PCs based on ourselves.

  24. anarchist said: "Players seem to compare their stats to other player characters, not to NPCs or to the range of possible results."

    That's a really nice, succinct point.

  25. [b]OZWALDO D. EMINADDI[/b]
    [i]STATS: AL: LG; AC10; MV 6; middle aged zero-level Halfling ; hp 2; THACO 20; #AT 1; D: nil (unarmed); S8, D10, C11, I19, W19, Ch 9; ML 17; no spells or Psionics.[/i]

    I offer Ozwaldo D. Eminaddi created by Roger E. Moore in the mid 90s. He had no problem with creating NPCs with stats outside the possible.

  26. The chargen mod system in Top Secret and later Gangbusters comes from Boot Hill - BH was also a big influence on firearms combat and damage in TS as well.

  27. @Delta:

    What's ML?

  28. A method my gang and I used for a while that we quite liked was to make a 3x3 grid. The rows were labelled str/con/dex and the columns int/wis/cha.

    You then rolled 3d6 for each box in the grid. For each stat you had to choose from the appropriate row or column, and of course once you used a given box you crossed it out. This gave a chance for better than average stats, but still a degree of randomness. You couldn't guarantee you would get the high scores exactly where you wanted them.

  29. @ anarchist...for shame youngling. Morale (ML) determined what it took for your NPCs to cut and run when the knives were out.