I don't like point buy for a few reasons:

• 3d6 down the line or 4d6 drop the lowest is very easy to remember and goes quickly compared to memorizing the advancement rates and options.
• If you can take disadvantages or lower stats one place to boost them in another, it encourages min-maxing and strains the boundaries of the math.
• The idea of designing the intrinsic attributes of a character just feels wrong to me. Random generation was great partially because it helped immerse players and enforce some of the worldbuilding through the experience players see at the table. A player knows Paladins are rare because it's hard to meet the requirements. With point-buy, it's impossible to convey anything about the world or even how I should think about the default stats of characters in the world.

In a game where stats matter less (AD&D 2e and prior), random generation was fine. From 3e onward, good stats can boost your success rate in your core competencies by 20-25% and the unified mechanics meant that instead of getting carved out benefits under select circumstances, a high stat gave very broad bonuses to attacks, skills, saving throws, and other stuff across the board. One of the problems I see in modern D&D design is that attributes matter a ton but designers can't make assumptions about what attributes the characters have (since any generation method might be used). A baseline swing of -10% to +20-25% chance of success on any given roll is too big to design around.

I personally use 4d6 drop the lowest when I run D&D. The worst thing about random generation in modern D&D is that you can end up with either a godly or a borderline unusable character.

What I'd probably do if I were to write my own generation method: roll dice and consult a table to get your attribute array (arrange scores as desired). The table is a bell curve - results towards the center are more evenly distributed across all attributes, while results towards the extremes look more like a min-maxer's stat block. This has the world-building effect of showing how attributes distribute unevenly, but it also means you can design adventures targeting a stat block in the center of the curve. It's also "balanced" in a sense because you can ensure the modifiers ultimately sum to the same value across the stat block. Players, especially new ones, also don't need to think too hard about their character's attributes, they'll get a functional character at the end of the process.

As an example: Top of the curve is something like 12,12,11,10,10,8 (+1 total in modern D&D), and one extreme end of the curve is maybe 18,12,10,9,8,6 (still +1 total). The numbers are just for example, you could scale the total expected attribute bonus to match the genre or game, but I think this blends the strengths of random generation and the standard array method. If we want a process more like 3d6-down-the-line, we can skip the arrange step and have the table specify the attributes in exact order. If that feels too random and we want to let people pick their character class first and foremost but still skip the arrange step, then a table can be provided per-class. That might even streamline the character creation process since all your info is on one page.

All of that is still slower than a 3d6/4d6 roll, but it guarantees a usable character at the end, simplifies math, avoids min-maxing incentives, and conveys a little bit of worldbuilding information.

Quote from: Eirikrautha on December 16, 2023, 09:43:26 PM

Quote from: hedgehobbit on December 16, 2023, 07:52:20 PM
The one thing that random character generation apologists tend to ignore is that player get an infinite number of rerolls. Every time a character dies, you can roll another, so you can effectively churn through characters until you have an above average roll. So any class or race balanced by die roll requirements, such as a Paladin, are really just a matter of persistence.

With a point buy system there is no advantage to a reroll-by-death so there is no incentive to kill off a loser character. IME this alone leads to better play.

This is one of those complaints that ranks right up there with old school tables being exclusionary and min-maxers can't roleplay.  I hear it from time to time, but I've never seen it in 40+ years of gaming.  Sounds like something theoretical that people make up to support their preferences and not an actual observation at the table...

Yeah, far more common in my experience is the GM just not giving a shit what the players rolled because he can always just adjust things behind the screen and feel no guilt at all about turning that "yup, I rolled straight 18s and max hp" PC into paste.

By contrast, I also don't know of a single GM who uses random rolls that uses 3d6 in order or even 4d6 drop lowest and arrange.

Most typical from my experience when the GM isn't just using the honor system is either "4d6, drop lowest, reroll 1s and 2s until they're not 1s or 2s, place in any order" or occasionally "roll 1d10+8 six times and place in any order."

By far the most common though in my experience is to use a point buy or array (with the bar none most common array I've seen from GMs being 18, 17, 16, 15, 14, 13; with my first exposure to that being Boy Scout summer camp in the 80's and continuing up through today).

Quote from: Old Aegidius on December 17, 2023, 07:08:10 AM
I personally use 4d6 drop the lowest when I run D&D. The worst thing about random generation in modern D&D is that you can end up with either a godly or a borderline unusable character.

What I'd probably do if I were to write my own generation method: roll dice and consult a table to get your attribute array (arrange scores as desired). The table is a bell curve - results towards the center are more evenly distributed across all attributes, while results towards the extremes look more like a min-maxer's stat block. This has the world-building effect of showing how attributes distribute unevenly, but it also means you can design adventures targeting a stat block in the center of the curve. It's also "balanced" in a sense because you can ensure the modifiers ultimately sum to the same value across the stat block. Players, especially new ones, also don't need to think too hard about their character's attributes, they'll get a functional character at the end of the process.

As an example: Top of the curve is something like 12,12,11,10,10,8 (+1 total in modern D&D), and one extreme end of the curve is maybe 18,12,10,9,8,6 (still +1 total). The numbers are just for example, you could scale the total expected attribute bonus to match the genre or game, but I think this blends the strengths of random generation and the standard array method. If we want a process more like 3d6-down-the-line, we can skip the arrange step and have the table specify the attributes in exact order. If that feels too random and we want to let people pick their character class first and foremost but still skip the arrange step, then a table can be provided per-class. That might even streamline the character creation process since all your info is on one page.

All of that is still slower than a 3d6/4d6 roll, but it guarantees a usable character at the end, simplifies math, avoids min-maxing incentives, and conveys a little bit of worldbuilding information.

I think this is an interesting idea, especially when you consider that if a Fighter's strongest stat is STR, someone with a wimpy 8 wouldn't even bother becoming a Fighter. As for the slow part, I don't think this would be too much slower than random generation.

One method I adapted from the Dragon Age RPG was rolling down the line, but once you're done rolling you can swap one score with another. That way if you got a 17 INT and an 8 STR, but you really had your heart set on playing a Fighter, you can swap those scores. In a lot of ways you can still get random scores, but you have enough control to play what you want.

In classic D&D random attributes worked fine because attributes were not so connected to static bonuses that affected every aspect of play. You could play a cleric with a 12 Wisdom and be perfectly capable. Over time more and more importance was heaped on stats until high stats became mandatory for success in anything. This escalation in stat importance was mapped closely with the diminished importance of player input mattering to the outcome of play. The "game" becoming just an exercise in play acting as all matters relating to successful outcomes were transferred to the character sheet at that point.

Quote from: Socratic-DM on December 16, 2023, 04:52:34 PM
...character creation can take forever depending on the concept...
...it can sometimes be prone to jank or powergaming depending on the group...

A point buy system is prone to jank or powergaming for sure, but I think that's dependent on the developers, not the groups. If you have a system with loose stat mods you probably don't get anything out of point buy, and that's how B/X or Worlds Without Numbers is set up.  WWN is sure to give you a 14 (the smallest stat possible for a +1 modifier) to at least one stat.  If you have a system in the 3.0 style, where each point or second point has ramifications to a decent number of rolls, then point buy becomes more important. 

If you run 5e, not only do you have the 3.0 stat progression which almost demands a point buy, you also have the ability to, as you level, either gain stats or gain feats.  Meaning if you start with the stats you need, you can spam feats- an incredibly huge advantage.

Basically, if your system is tight and demands balance, then you need point buy, and the attribute spread is about your character's interaction in the game world with the rules.  A loose stat system, as almost the entire OSR is all about, is about mapping your character's stats in a realistic way.

Which one is better?  I think the older method is better, overall, but they both have their merits.

Quote from: Grognard GM on December 16, 2023, 05:09:30 PM
I hate random char gen.... 

I'm not even totally against a mixture of the two.  ... with initial stats being a modifier plus random roll. It's kind of a controlled randomness.

So you like it just fine, your problem is with the variance of it, because something like "3d6 down the line" or "4d6 drop lowest" will generate really wild ass nonsense like 10% of the time, and as such many tables will often have wild ass nonsense at them.  I'm glad you got in with a first post on this because in most of these conversations, the position of "your point pool is N+1d6" is buried deep, despite being something I've seen time to time for years.  The player rolling low on this still has a viable character with whatever the max in his favorite stat, and the one rolling high can kinda break a bit of the boundaries without being wild like starting with three 16s or two 17s or whatever.

I think the biggest difference between point buy and random rolls is, do you want to play the preconceived character in your head or do you want to play the character the dice give you?  Where "you" is really what the group and the GM want, since it should generally be the same for everyone at the table.  There's nothing inherently wrong with either approach.  They can even play very similar in some cases. I think the biggest difference are on the edges. 

For example, if you've got a group of players that are constantly wanting to try new things, always have half a dozen ideas in their heads of what they want to play next, and tend to use mood, other players ideas, etc. to introduce almost some randomness to their character concept--then point buy is great.  The randomness is all in the concept, and then the point buy is used to realize the concept. Such players tend to not push the boundaries of the system they are using.  OTOH, if you have players that are quite willing to explore some new ideas giving a prod, but they need that prod to get started, then random is better.  Not only are they likely to get out of ruts, they are likely do so in ways that even the other players with half a dozen ideas might not have tried.

The less likely death is to occur, the more useful point buy is. Likewise, the more likely death is to occur, the more useful random gen is.  This isn't only in the negative sense of getting your lousy generated character killed, though even that has a positive aspect. I've never seen someone suicide a character with bad stats.  I have seen multiple times a character with bad random stats get played aggressively on the grounds of "make something of this character quickly or die trying."  This has interesting side effects on party dynamics, and is seldom the kind of notion you'll see in a crafted concept with point buy.  Moreover, the statistical likelihood of death also relates directly to the time you are likely to spend with the character.  A rough edge that you can't do much about is more palatable if you know you may not keep that character forever, and if you do happen to live, it even takes on kind of a badge of honor.

Note that random generation and point buy are only one way in which these dynamics can manifest.  Random hit points at low level is arguably a bigger deal in some systems.  Systems with life paths that are partially random have their own aspects. 

Finally, I think what can happen after character generation is more important than the first pass.  It's easier in GURPS, for example, to fix a mistake in the initial character, because you just spend more points. That would be true even if you used some house ruled random generation to start.  It's also easier in GURPS for characters to become very similar as they get points, unless this is policed somehow by the group.  Whereas something like early D&D will be a lot more palatable to many that otherwise wouldn't like it, if there were more ways to improve the character's abilities as they went (never mind those that don't care, since the level increase is much more important that the stats).

For all of these reasons, I prefer random generation at start, with some key, meaningful player choices factored in, to give a mix of preconceived concept and rolling with what the randomness gives. Then I also prefer that the system have built into it (math, model, etc.) similar meaningful chances to improve on the random bits as the game progresses. 

My own preference is for a random roll with a relatively bounded set of results. I've started running my OSR-like games using 1d6+1d4+6 as the attribute roll, technically a range of 8-16, but most rolls come out in the 9-13 range, and then I let people boost an attribute every 3 levels or so. I wouldn't use that for 3rd-5th edition though. Like others said, those games presume higher attribute modifiers.

If you're not going to roll attributes, then I feel like just using a "standard array" makes more sense than point buy. It's faster and lets you to control the amount of min-maxing the players can do.

I run GURPS 4/e - for at least 20 years now - got no problem with 'point buy' systems.

However, I am not 'great' at math and don't expect my players to be either.
With GURPS what has been a minor godsend is that program called 'GURPS Character Assistant ' (GCA).
With the GCA I can choose from a template in there and then if a player is in the same room with me or sitting near me at the screen they can customize the character to their liking when its first created.

As to the other issues like 'power gaming' or unbalanced characters - I nip that in the bud from the very beginning. The players are told that the skills assortment should be reasonable and plausible, anything unusual has to be justified by an interesting back story. Of course with GURPs an interestying back story gives you disadvantages, quirks, and advantages that make the character more interesting.

Heck, take Captain Christopher Pike - "Likes To Cook" makes a mildly interesting quirk on a character sheet.

-Ed C.

Every design decision is about pros and cons. For every fault that can be found (real or imagined) in Point Buy something else (sometimes similar) can be pointed out of Random Generation. It's all a matter of implementation and what you want out of the game.

If you want customization and fine tuning your character, there's no amount "but what about teh powergaemrz!" Or "character creation takes too long!" is gonna change the fact that random generation sucks for that purpose. Plus there are ways around those issues, some of which have already been brought up, such as templates, or the GM putting their big boy pants on and saying "NO!" to certain things during character creation. So some of these cons of point buy are really just "cons" if you let them be, or if the system is designed badly.

I like point buy in principle, but in practice I find it doesn't work too well.

I've experienced two problems. The first is once you figure it out, there's a right way and a wrong way to design an effective character. Given the same point budget, someone who knows what they are doing will have a much better character than someone who is just trying to realize the concept. The second is that it's sometimes very difficult to realize a concept within a budget, and this isn't necessarily a concept that is over powered. Just that it ends up costing too many points to have mechanical fiddly bits that represent the fiction you want to convey.

I haven't come across a single point buy system, and I've looked at many, that satisfactorily solves either, let alone both, of these problems.

Quote from: migo on December 17, 2023, 04:32:45 PM
I like point buy in principle, but in practice I find it doesn't work too well.

I've experienced two problems. The first is once you figure it out, there's a right way and a wrong way to design an effective character. Given the same point budget, someone who knows what they are doing will have a much better character than someone who is just trying to realize the concept. The second is that it's sometimes very difficult to realize a concept within a budget, and this isn't necessarily a concept that is over powered. Just that it ends up costing too many points to have mechanical fiddly bits that represent the fiction you want to convey.

I haven't come across a single point buy system, and I've looked at many, that satisfactorily solves either, let alone both, of these problems.

I haven't come across a single random generation system that satisfactorily solves either of these problems either. Which is why I mentioned in my post above that "For every fault that can be found (real or imagined) in Point Buy something else (sometimes similar) can be pointed out of Random Generation."

Every single time that threads come up where people nitpick some aspect of point buy systems. The issues brought up are almost invariably: things that also apply to random generation (or just about EVERY system out there), or stuff that's a matter of inevitable give & take/pros & cons of game design (where something's gotta give regardless, so you have to make up your mind about what you want most: the design option, or to avoid the issue you don't like. But you can't have both).

Quote from: Socratic-DM on December 16, 2023, 04:52:34 PM
the two biggest complaints I personally have is that they 1. character creation can take forever depending on the concept, 2. it can sometimes be prone to jank or powergaming depending on the group.

Yes, that plus player regret. If you choose X and Y, then when X turns out to be useless and Z turns out to be needed you regret it. In its most benign form this is simple player disappointment. But it can turn nasty, where the player gets annoyed with the system or the DM, "but you didn't tell me X was useless and I needed Z!" Whereas if they rolled it up, they just shrug and move on, it's just the character generation of rolling minimum damage in every hit in a combat - shit happens.

For newbies, your two points also combine with my third. If someone's new to our game group, especially if they're new to gaming, we want to things to be accessible. Rolling for everything is accessible, having to choose is not. That's because the choices require knowledge of the game system and/or campaign world - but the newbie, by definition, does not have this knowledge. So if you want to get new people in your game group and expand the hobby generally, you want systems which lean towards random roll.

Remember too that even if you have a well-established game group which somehow never sees anyone move house, change careers, have children and so on and thus have to move on, you also want their gaming to be acceptable to the people in their life. It's hard for Jen to come game if her husband Bob thinks it's demonic, or overly-complex nonsense, or the books cost more than their monthly groceries, or whatever. He's going to hassle her every week as the game session comes up, and she's going to start associating her game session with marital troubles, and eventually stop coming. But if it's obviously harmless, accessible and cheap, then Bob will at least put up with it as, as he sees it, an eccentricity of his otherwise perfect wife.

And so the best gaming is one which is accessible and understandable not only to actual gamers, but lots of other people, too. Random rolls is not always but tends to be more accessible and understandable.

So where someone like Pundit or me implies that point-buy is inherently bad, that's the unspoken assumption there: we want to expand the hobby and make newbies and current gamers both feel welcome and comfortable.

Quote from: VisionStorm on December 17, 2023, 05:59:22 PM

Quote from: migo on December 17, 2023, 04:32:45 PM
I like point buy in principle, but in practice I find it doesn't work too well.

I've experienced two problems. The first is once you figure it out, there's a right way and a wrong way to design an effective character. Given the same point budget, someone who knows what they are doing will have a much better character than someone who is just trying to realize the concept. The second is that it's sometimes very difficult to realize a concept within a budget, and this isn't necessarily a concept that is over powered. Just that it ends up costing too many points to have mechanical fiddly bits that represent the fiction you want to convey.

I haven't come across a single point buy system, and I've looked at many, that satisfactorily solves either, let alone both, of these problems.

I haven't come across a single random generation system that satisfactorily solves either of these problems either. Which is why I mentioned in my post above that "For every fault that can be found (real or imagined) in Point Buy something else (sometimes similar) can be pointed out of Random Generation."

Every single time that threads come up where people nitpick some aspect of point buy systems. The issues brought up are almost invariably: things that also apply to random generation (or just about EVERY system out there), or stuff that's a matter of inevitable give & take/pros & cons of game design (where something's gotta give regardless, so you have to make up your mind about what you want most: the design option, or to avoid the issue you don't like. But you can't have both).

With random generation, you're not expecting to be able to make a character according to your wishes. So the second point is automatically a non-issue. The first point, yeah, you can have character imbalance with random generation, but there you only have one issue to solve, not two.

Point buy for Dungeons and Dragons is bad because the system is built on the random method, and characters with powerful stats will be rare due to dice odds. Which is an assinie approach because players generally want a competent character, and will wheedle, cajole, and sometimes outright cheat to get a "good" character. That's why I tend to favor the Stat Array method in most cases. Here's your stats, now shut up and play.
Point buy systems in other games have their issues mentioned in this thread, but also have the advantage of being accounted for from the beginning. If you're playing Champions, you know what you're getting into, and the rulebooks mention power gaming and how the GM should keep that in mind during character creation.

Quote from: Exploderwizard on December 17, 2023, 11:10:54 AM
In classic D&D random attributes worked fine because attributes were not so connected to static bonuses that affected every aspect of play. You could play a cleric with a 12 Wisdom and be perfectly capable. Over time more and more importance was heaped on stats until high stats became mandatory for success in anything. This escalation in stat importance was mapped closely with the diminished importance of player input mattering to the outcome of play. The "game" becoming just an exercise in play acting as all matters relating to successful outcomes were transferred to the character sheet at that point.

Insightful. Now for those who don't know:

In AD&D1e, you need Wisdom 9 to be a cleric at all. You need Wisdom 17 to cast 6th level spells. But Wisdom 9-16 can cast up to 5th level spells, once the right class level has been achieved. Clerics don't get access to 6th level spells until 11th level, which requires 675,001 experience points - so the spell level limit wouldn't be an issue for quite a while. And it would of course not stop them, as they levelled up, getting more spells each day of 5th and lower levels, more hit points, better saving throws and so on.

Wisdom 9 gave you a 20% chance of spell failure; 10 a 15%, 11 a 10% and 12 a 5%, with 13+ no chance. 13+ also gave you extra spells to cast each day. So while 13+ was a big advantage (casting two spells at first level instead of one helps a lot!) it wasn't crucial.

Intelligence 9 was likewise necessary to be a magic-user. Since 10 was needed for 5th level spells, 12 for 6th, 14 for 7th and so on, an Int 9 character would most crucially feel this limit at 9th level, where 5th level spells were normally needed. Learning new spells for the Int 9 MU would be a slow process, with only a 35% chance of understanding a new one encountered, rising to 45% at Int 10-12, 55% at 13-14 and so on. The Int 9 MU's minimum number of spells known each level is only 4, and their maximum 6.

Strength 9 is needed for a Fighter. Damage bonus of +1 doesn't appear until Str 16, and a to-hit bonus of +1 at Str 17. But even with 18/xx Strength being common, the character's level was always a bigger factor. The best possible bonus is with 18/00 strength, giving +3 to hit and +6 damage. But a fighter gets effectively a +1 to hit each level (the chart has +2 every 2 levels, but we can assume that was more to keep the chart a sensible size, and simply interpolate it as +1 each level). Damage does not increase with level, but while the Strength to-hit bonus is limited to +3, the level to-hit bonus is not. And so in just a few levels, the level of the Fighter is a far bigger factor than their Strength.

Likewise Thieves and Dexterity. There are some maluses to Thief abilities at Dex 12 and below, and bonuses from 16 on. But their abilities go up on average 5% each level, and so not many levels pass before Dex becomes minimal in importance. The greater possible effect would be the missile weapon to-hit bonus coming in from Dex 16, and the AC bonus from Dex 15. But again, the to-hit bonus from Dex will be eclipsed by that from going up levels (+2 by levels 5-8), and as for AC, a thief should not be getting themselves into direct combats in any event, their job is to sneak and ambush.

And so we see that in AD&D 1st edition, attributes are of low importance. Level will be a greater factor, and in order to achieve those levels, player wits.