Random Attribute Scores

[Sean]

OK. It sort of embarrasses me that I'm still thinking about this, but I am. You can't always be one of the smart kids if you want to make progress.

There were three things I really liked about old-school 'roll your attribute scores' character generation:

1) No muss, no fuss. It took under 1 minute to generate the bare bones of your character.

2) Some great ideas came out of free associating on those numbers, some of which were ideas I wouldn't have come up with myself. They were a spur to creativity on my part when it worked well (see below).

3) Taking players out of their comfort zone and challenging them to do something a little new. Is this a virtue? In some cases, when it led to outcomes of type 2, it was. In the majority of cases, though I think it wasn't. Which leads to my three problems with random attribute score generation:

A) You could roll up an individual character who was seriously teh suck. Everyone knows this one and there are a million ways to fix it now, but it bears mentioning.

B) Group problems. Even once you get the process fixed for individual characters, you sometimes get a group with e.g. all high strength and low intelligence, or the like. This gets in the way of people feeling like their characters are individuals, and also leads to the overshadowing problem, where one guy is a little to a lot better than the other guy at almost everything.

C) Player/character mismatch problems. So you roll up a set of stats that could be a very plausible dwarf fighter. Well, that's great, but what if you hate dwarves? This is the converse of 3 and to some degree 2 above: random character generation is great when you're up for it and the dice cooperate a little, but sometimes they just don't.


The challenge for this thread, if anyone wants to take it, is to find a way to fix (C) that doesn't involve giving up total or near-total randomness in the generation process. Stated thus, as I see it that may be flatly impossible, but maybe near-solutions could be offered. (Here's a near-solution: keep rolling up a character until you get one you like. This is actually the way a hell of a lot of random systems, and not just for attributes, wind up getting used in practice. Problem: is there any way to make that functional, or to work with this kind of 'choose the random thing that inspires you' idea that could be reasonably implemented in a game?)

I wouldn't mind suggestions to fixing B as well, but I've decided that I can live with that problem (in my application for this thread) if I can get a little more of a handle on C.

[Paul Czege]

Each player rolls up and completely creates a character. Then negotiate as a group who will play which character. You earn experience points for your own play, and for the play by the user of the character you created.

(For added gamism, the negotiation phase can include bidding for pick order: first pick takes a 50% experience point penalty, second pick takes a 45% penalty, etc.)

Paul

[Mike Holmes]

Paul beat me to that one. But I do think it's a "near-solution."

How about if it wasn't the stats that were completly randomly generated, but to whom the stats already generated were given that was random. OK, bad sentence. Example:

1. There are three players, so three characters.
2. They each need a strength score, so the system provides the three strength scores somehow, say (using the familiar D&D range) a 5, an 11, and a 16. This can be somewhat randomized itself, but the scores overall would "balance out". IOW, the even distribution that I'm giving would not have occured out of random chance, but by design.
3. Then you roll to see who gets each score.
4. For the next stat, the character who got the high score can't get it again - he can only get the medium or low score. Basically, there's some sort of round robin mechanism that prevents any one character from getting all of the highest abilities.

This solves several of the problems. No character will be "teh suk" (solves A). Each character will automatically be different from the others (solves B). When done, then you negotiate to see who plays what. Given a good distribution of players, each player should get something they can live with (solves C).

In fact, if you want, you can have each player mess with the distributions before hand in case the group is feeling more like playing all fighters or something. But this starts messing with randomness again (like the D&D rules where you select Barbarian and roll 6D6 and take the best three). As you say, however, this is insoluble. Basically the contradiction is that you can't roll characters randomly, and then not have any unwanted options come up at least occasionally.

You can mitigate, however. For instance, you could have metagame currency useful for effectiveness in-game after chargen or somesuch, that could be burnt in chargen to allow for "rerolls" (and allow players to go back to previously rolled characters, too). That gives the player an incentive to take his random character (especially if the metagame points are really important), but allow for outs with characters they really don't want to play.

There is a problem with this, too, however, which is that some player will manage somehow to roll up a character he doesn't want to play 5 (or whatever) times in a row, and burn through all of his points, still have to play a character he doesn't like, and have to do it without any metagame points.

To some extent, I think that you can mitigate this by allowing for modification after the fact in some ways. I mean, why would a particular set of D&D stats indicate a Dwarf automatically? Why couldn't the same stats be used for a human. If you allow enough maleability after the fact, and use a method like the above to ensure no suk characters, then I think it might work.

Mike

[Doug Ruff]

Both of the above look good - and interesting - to me. Here's another:

(1) Create a fairly large number of template characters - for a D&D or Fantasy Heartbreaker system, you could have several warrior templates (burly guy, duellist, soldier, knight, archer etc etc.), same for rogues, wizards, priests, experts, whatever.

Each template has a small number of options, so they're not identikit (the templates in Feng Shui are a good example of what I'm thinking of. Or each template has 2 or 3 different pre-gen sets of stats and abilities to pick from.

(2) Each player picks 6 templates that they would like to play - they don't have to be similar types, just 6 they would enjoy playing.

(3) For each player, roll a die to see which template that player gets.

Upside of this method is that no-one gets a sucky character, and you can apply as much (or as little) 'balance' in advance.

The downside is the need for a stack of pre-gen templates. Some games already have 'rogues gallery' type supplements for this, then again, some people like chargen for the hell of it, so this may not be a chore.

Doug

[Lxndr]

The idea I've suggested several times to D&Ders, but which I've never tried or done myself, but I will eventually put into a game design if I find a place that warrants it:

Take the # of dice that'd normally be given to the group (for D&D, let's say 18) and divide it evenly among the players.  So with 3 players, each rolls six dice (if it's uneven, round up, and then drop the lowest).  Those dice numbers are then tossed into the center of the table.

Players then take those dice, in any groups of three they want, and assign them to their characteristics.  So, every player gets the exact same pool of dice to draw from, but everyone rolled, and thus contributed.

This might possibly not random enough for your C, since the players assign the dice after they're rolled.  But it's still an interesting idea, and if it's not random enough for you, perhaps it might spark your own thoughts on the matter.

[M. J. Young]

Actually, I think AD&D solved this one pretty well with the Method V rolling system in Unearthed Arcana (which I've expanded somewhat http://www.mjyoung.net/dungeon/char/step005a.html">here. The http://www.mjyoung.net/dungeon/char/step005.html">system for creating Gully Dwarfs in Dragonlance Adventures has similar marks.

In essence, the player decides what kind of character he wants to play, either singly or in cooperation with the others (most of the players I've known when they join a new game start by asking what kinds of characters are already in the group and what the group thinks it needs). The method of rolling the scores is still random, but it is differently random for each score, depending on what kind of character is intended. In the Method V system, the player is going to take the best three of nine rolled dice for the one score that is thought to be the most important for the kind of character he wants to be, and then the best three of eight for the next score, and down until the seventh and least important of the scores rolls three dice and takes whatever is rolled. The result is that the important scores for that character type have a very strong probability of being very high, while those scores that are not important to what the character is supposed to do have lower probabilities of the same.

In one of my Game Ideas Unlimited articles, http://www.gamingoutpost.com/GL/index.cfm?action=ShowProduct&CategoryID=54411&ProductID=71772&publisherid=54849">Negative Points, I suggested a method of randomized scores for a game in which scores would range from one to twenty.

...the player is allowed to roll, once each:

  1. d20 (1 to 20, linear)
  2. 3d6 (3 to 18, fairly strong tendency toward 10 or 11)
  3. 3d4 (3 to 12, fairly strong tendency toward 7 or 8)
  4. 4d4 (4 to 16, very strong tendency toward 10)
  5. 2d10 (2 to 20, weak tendency toward 10 or 11)
  6. 2d4+12 (14 to 20, weak tendency toward 15 or 16)

and arrange them in any order desired.  This gives a lot of different curves, assuring that at least one score will be at least 14, at least one will be not greater than 12, only one could be below 2, and so forth.  It is still very random, but it has a bit more stability in the differences between characters.  Clearly you will not get one character with all twenties and one with all twos.

Randomized systems don't have to be as random as some are. The point of the randomizer is to create variety and equity. That can be achieved with less chance of incredibly low numbers and more player control over where the strong numbers land.

I hope this helps.

--M. J. Young

[John Kim]

C) Player/character mismatch problems. So you roll up a set of stats that could be a very plausible dwarf fighter. Well, that's great, but what if you hate dwarves? This is the converse of 3 and to some degree 2 above: random character generation is great when you're up for it and the dice cooperate a little, but sometimes they just don't.

The challenge for this thread, if anyone wants to take it, is to find a way to fix (C) that doesn't involve giving up total or near-total randomness in the generation process. Stated thus, as I see it that may be flatly impossible, but maybe near-solutions could be offered. (Here's a near-solution: keep rolling up a character until you get one you like. This is actually the way a hell of a lot of random systems, and not just for attributes, wind up getting used in practice. Problem: is there any way to make that functional, or to work with this kind of 'choose the random thing that inspires you' idea that could be reasonably implemented in a game?)

OK -- maybe this is too simple, but what about just a limited number of re-rolling, like say three max?  I think the main problem with re-rolling is that it takes a long time, not that there's anything wrong with the method.  I mean, by having randomness, you want to force players to get things they wouldn't choose.  For point C, you just want to mitigate that a little bit to avoid highly ill-fitting options.  

Just speaking from experience, in preparation for an upcoming (unfortunately delayed) HarnMaster campaign, I ended up rolling three characters.  The GM had put in house rules to beef up characters by putting in some choices at different points in the character generation process.  I instead asked to not use the house rules but instead roll straight but do it three times.  This gave me three more complete random characters, which were inspirational.

[FzGhouL]

I personally like having basic attributes, then letting the characters add one and roll for a random one added.

Unfortunantly, this only works in systems where 1 Attribute is a big deal, you could however up the amount.

I just don't like when one character gets a badass roll and the other gets god awful hehe.

[Sean]

Thanks for the replies! These are all good options.

The character-trading one is especially useful for what I'm up to; I'm going to use it. . Also, just allowing a 'free' reroll if it's hopeless might be functional as long as you indicated that character creation was a group process. Though I can think of some kid groups who would contract to break this together. Maybe that's their perogative.

I really like the bidding idea too, but that winds up being more like a point buy/design system, as do the multi-sized die pools. In the D&D context, swaps and rerolls (one of each is good) are a nice way to mix design and randomness.

Allowing players to swap with each other could form an interesting currency. Hmm...interested to hear more. Actually, this general subject isn't really all that lame - should have kept my embarrassment to myself.

[NN]

But, MJ, the side effect of Method V is that the statistics are much, much  higher than in the normal 3 dice or 3 dice out of 4.

Surely the easy fix to C) is to roll the stats randomly, but be allowed to choose the order.

Also, if its a niche-class type of game like D&D, then the players can take it in turns to order their stats and choose the niche - with the worst set of stats going first..

[kenjib]

Here is a method for D&D type stats that addresses all three of these issues while still preserving a level of randomization that gives some of the benefits you ascribe to free association.

Step 1:  First you roll 3d6 for each score, straight down the line, in order.
Step 2:  You may switch any two scores.
Step 3:  Calculate the total value of your rolled scores using the 3rd edition point-buy system* (any value 8 or less counts as 0).  If your total is less than 28, you get the difference in points that you can spend to increase your scores according to the point-buy cost scale.  However, you always get a minimum of 4 points to spend.  So, if your character's stats add up to 21 points, you get 7 more to spend.  If they add up to 26, you get 4 more to spend, even though you are only 2 points under 28.

*The 3rd edition point buy system is a system where, roughly, stats become increasingly expensive to buy the higher they get following an exponential type of curve.

Problem A is addressed in step 3.  Problems B and C are addressed in steps 2 and 3.

You can pretty much always create the basic type of character you want, but there will almost always be something unexpected in his stats too.  I'm not sure if this is still random enough to satisfy your criteria, as it straddles the line pretty solidly between stat buy and stat randomization.

[daMoose_Neo]

To toss another idea out on the field:
1) Choose what you want w/ a base set of values (thus the Mage/Sorcerer will have naturally lower strength to the fighter while the fighter will have a lower intelligence compared to the Mage/Sorc)
2) Roll a set of random die appropriate to the system for the number of stats you have . D&D, cap the max starting values at 8 (6?), roll one set of 2d6 for the 6 stats.
3) Plug & Play - plug the set of values where you want (so ideally a Mage would put his higher roll into Int, while a fighter would beef up Str, and then possibly stick their next highest value in their lowest base stat to shore up some weakness)

Really like the ideas of group gen and then seeing who gets what, would make for some interesting and unexpected games ^_^

[Mike Holmes]

But, MJ, the side effect of Method V is that the statistics are much, much  higher than in the normal 3 dice or 3 dice out of 4.

Well, the real problem with this method is that it voids one of the criteria for the system which is that it actually produce random types of characters? The UA system will produce the same character over and over, with what amount to minor variations only. Sean's demand, if I have it correctly, is to produce random characters such that the player may be inspired to stretch out and play something that they might not have otherwise. If you start from the assumption of a character class, for instance, and force the randomization to fit that, then you're not meeting his criteria, IIUC.

Same problem with moose's solution, of course. Ken's, is a very compromise position - it might force you to play something other than what you were thinking, but it certainly won't force you to play some specific thing. So it might be good.

Most of these things have the "flaw" that they might actually end up producing exactly what the player would want to play if given their druthers. Not sure if that's something to avoid or not. :-)

Mike

[Lee Short]

Sean,

Here's something I used to do for an old game.  Each player has about 10 character points.  These can be used to buy contacts, positions, cool equipment, wierd background, etc.  They can also be used while generating your character's stats.  You roll 3d6 for each stat in order (or you may simply take a result of 10 for any stat).  You may spend character points to:

-- roll 4d6 and keep the best 3 (must be declared before dice are rolled)
-- swap 2 stats
-- completely reroll a character you don't like
-- give +1 to a selected stat

What worked well about this was that it was all up to the player -- but the more the player just went with what the dice gave him, the more character points he had to buy cool stuff later.

[WiredNavi]

My tired brain says:

If you're looking for A and B and not C, then you might think about trying a random generation system for 'bare bones' that is NOT based on stats.  For instance, tables which randomly generate some kind of character aspect ('haughty', 'gambler', 'overconfident', 'loyal', etc).  This gives you the bare bones of a character concept without necessitating many things you might find distasteful - you could allow a reroll or two to avoid aspects you don't want to play or something.  Then come up with a character concept that incorporates those aspects and stat accordingly via whatever method you feel is best.

This might give you what you want - Quick, bare bones concepts, plus spurs for new and not-previously-considered characters, without the necessity of having the wrong numbers in a Gamist sense for what you want to play.

Sorry if this is marginally less than coherent.