D&D 5E Point Buy vs Rolling for Stats

[Lanefan]

Why? The results of point-buy, the standard array, and the array of the Commoner NPC are all possible results of taking the highest three dice of 4d6.

True.

Why are the numbers more realistic when rolled?

Variability.

People in the real world are not "balanced" anywhere near as evenly as we would be if we were all using the same stat array. The range between best and worst simply isn't big enough, and we don't all perfectly trade off at being just as good at one thing and just as bad at another.

Ditto for if we were all built by point-buy - there'd be a bit more variability but nowhere near enough range. Take intelligence by itself: people exist with IQs of 30, and of 180 or more, across a bell curve that peaks somewhere in the 110 range I think. Now there's no denying that this real-world bell curve is a bit tighter than rolling 3d6 produces, but there's also no denying its existence and there's also no denying that the 3d6 model is a reasonably elegant, if imperfect, attempt to model it.

The game tries to model that same bell curve across six stats instead of just one; in all cases the resulting bell curve is too loose, but in all cases the model sort of works as intended.

Now, in reply to [MENTION=22779]Hussar[/MENTION] and [MENTION=284]Caliban[/MENTION] and others who say the general populace doesn't have stats at all:

Somewhere deep in the Canadian wilderness is a tree, one among millions. No living person has seen this tree except maybe from the air as part of a huge sprawling forest, and no specific records have ever been made of it. Does it still have a height? A girth? An age? A species? Of course it does! Even though we don't know the specifics, all those things exist and should anyone ever bother to go and measure them actual results would be forthcoming.

The same holds true of unseen elements in any kind of game world that exists beyond the PCs sight - and even what exists within the PCs sight. As the party enter Neverwinter the streets are teeming with people, and the DM maybe even mentions this. Does each person in that crowd have stats independent of each other person? A height, weight, and age? Of course they do! Even though we don't know the specifics, all those things exist - just like the tree in the forest - and should anyone ever bother to determine them actual results would be forthcoming. But how to determine them, is the problem.

With trees, we can look at a whole bunch of individual trees of a given species and from there pretty well model the rest - we know the averages and expected variances, we mostly know the extremes, and we can even take external factors into account e.g. climate to determine what might be expected in a given place. With people in a game world the game gives us the extremes (3 and 18 for Humans, normally), it gives us the averages (10.5), but it doesn't give us the expected variances nor any means of determining them, nor - with the exception of 1e's 3d6 - does it give us an expected "tightness" of the bell curve.

And while I agree with those who say that none of this is needed until it's needed, I disagree with the idea that none of it exists until it's needed. It all exists. We just haven't measured it; where in this particular case measuring it means rolling it up.

Put another way, if for whatever reason I determine that the third guard on the left has a 15 strength I've also just determined that said guard has always had a 15 strength - I just didn't happen to know about it.

Lan-"can't see the trees for the forest"-efan

 

[Oofta]

True.

Variability.

People in the real world are not "balanced" anywhere near as evenly as we would be if we were all using the same stat array. The range between best and worst simply isn't big enough, and we don't all perfectly trade off at being just as good at one thing and just as bad at another.

Ditto for if we were all built by point-buy - there'd be a bit more variability but nowhere near enough range. Take intelligence by itself: people exist with IQs of 30, and of 180 or more, across a bell curve that peaks somewhere in the 110 range I think. Now there's no denying that this real-world bell curve is a bit tighter than rolling 3d6 produces, but there's also no denying its existence and there's also no denying that the 3d6 model is a reasonably elegant, if imperfect, attempt to model it.

The game tries to model that same bell curve across six stats instead of just one; in all cases the resulting bell curve is too loose, but in all cases the model sort of works as intended.

Now, in reply to [MENTION=22779]Hussar[/MENTION] and [MENTION=284]Caliban[/MENTION] and others who say the general populace doesn't have stats at all:

Somewhere deep in the Canadian wilderness is a tree, one among millions. No living person has seen this tree except maybe from the air as part of a huge sprawling forest, and no specific records have ever been made of it. Does it still have a height? A girth? An age? A species? Of course it does! Even though we don't know the specifics, all those things exist and should anyone ever bother to go and measure them actual results would be forthcoming.

The same holds true of unseen elements in any kind of game world that exists beyond the PCs sight - and even what exists within the PCs sight. As the party enter Neverwinter the streets are teeming with people, and the DM maybe even mentions this. Does each person in that crowd have stats independent of each other person? A height, weight, and age? Of course they do! Even though we don't know the specifics, all those things exist - just like the tree in the forest - and should anyone ever bother to determine them actual results would be forthcoming. But how to determine them, is the problem.

With trees, we can look at a whole bunch of individual trees of a given species and from there pretty well model the rest - we know the averages and expected variances, we mostly know the extremes, and we can even take external factors into account e.g. climate to determine what might be expected in a given place. With people in a game world the game gives us the extremes (3 and 18 for Humans, normally), it gives us the averages (10.5), but it doesn't give us the expected variances nor any means of determining them, nor - with the exception of 1e's 3d6 - does it give us an expected "tightness" of the bell curve.

And while I agree with those who say that none of this is needed until it's needed, I disagree with the idea that none of it exists until it's needed. It all exists. We just haven't measured it; where in this particular case measuring it means rolling it up.

Put another way, if for whatever reason I determine that the third guard on the left has a 15 strength I've also just determined that said guard has always had a 15 strength - I just didn't happen to know about it.

Lan-"can't see the trees for the forest"-efan

First, I don't see cookie-cutter PCs in my games. I think you are making a mountain out of a molehill. Or a redwood out of a sapling. Something.

But what do you know about the variation of tree sizes? If you had to generate the forest, what specifications would you use for height, width number of leaves or needles?

If you don't know anything about trees other than what an average white pine looks like and the range of possible values you have no idea what your model to generate them should be. You could end up with trees that are 20 ft diameter but only a foot tall, or trees that are 200 ft tall but have trunks that are a foot in diameter. And how many foot tall trees should you have? How many are 200 ft tall or more?

Similar with ability scores. We know ability scores should vary around some average value, but we don't know by how much. How many people should have an 18? How many should have a 3? No clue. I just know that 1 in 216 should not have the absolute minimum possible for intelligence. I would say that any Int score below a 8 would be considered "slow" and anything below 6 would be considered handicapped (based on the mental capacity of an ogre). Should 5% of the population be mentally handicapped?

We know 3d6 is bad at modeling ability scores, that's all. We don't know what the distribution should be therefore it's pointless to justify your opinion based on something we cannot possibly measure.

 

[Caliban]

Now, in reply to @Hussar and @Caliban and others who say the general populace doesn't have stats at all:

I never said they don't have stats. I said you may not bother to generate them, and I said NPC stats don't matter (meaning "specifically in regard to PC stats"). NPC's use different rules than PC's, how they get their stats is irrelevant to how PC's get their stats.

 

[Lanefan]

First, I don't see cookie-cutter PCs in my games.

It's cookie-cutter NPCs I'm after in this case.

But what do you know about the variation of tree sizes? If you had to generate the forest, what specifications would you use for height, width number of leaves or needles?

The specifications I'd determined by measuring and statictically analysing a whole lot of trees that I could access, to model those I could not access.

Similar with ability scores. We know ability scores should vary around some average value, but we don't know by how much. How many people should have an 18? How many should have a 3? No clue. I just know that 1 in 216 should not have the absolute minimum possible for intelligence. I would say that any Int score below a 8 would be considered "slow" and anything below 6 would be considered handicapped (based on the mental capacity of an ogre). Should 5% of the population be mentally handicapped?

We know 3d6 is bad at modeling ability scores, that's all. We don't know what the distribution should be therefore it's pointless to justify your opinion based on something we cannot possibly measure.

Except we can measure it for one stat, more or less, using real-world IQ scores or equivalent; and we can thus know pretty much exactly what the distribution should be across a population. It's not a big stretch to then for game purposes apply that same distribution to the other five stats.

We agree that 3d6 gives too loose of a bell curve for any stat.

I think we agree that the distribution for any stat across a population follows a bell curve of some sort, though we could debate for years how to arrive at a means of randomly generating individual scores along that curve.

I hope we agree that those six bell curves (one for each stat) won't all nicely cancel out for every given individual; that some individuals will be genetically luckier than others.

Where we disagree is how any of this affects PC generation as opposed to modelling the population at large.

With me so far? Or, if not, where am I going wrong?

Lanefan

 

[Hussar]

Heh, so, the fact that 3e never, ever used randomly generated NPC's is irrelevant, but a single throw away line, THAT IS CONTRADICTED IN THE SAME SENTENCE, is? Remember, they SPECIFICALLY STATE THE ARRAY OF ELITE NPC'S AND STATE THAT NON-ELITE NPC'S HAVE LOWER STATS.

So, no, 3e NPC's are not generated by 3d6 in order, since that would result in significant numbers of them becoming Elites. And, ALL elite NPC's USE THE SAME STAT ARRAY.

Good grief [MENTION=23751]Maxperson[/MENTION], don't you get tired of cherry picking your points? Or do you see that one sentence that supports you and just mentally block out everything else.

Just to reiterate. Elite NPC's all have the same stat block, arranged to the most advantageous set. Non-Elite NPC's have lower stats because, well, if they had higher, then they would be elite and use the elite array. So, it's somewhat contradictory right in the same paragraph to say that all NPC's are 3d6 rolled. They aren't. They can't be since any 3d6 result that would exceed the Elite array would automatically have to shift that NPC to elite and use an identical array.

But, hey, keep on keeping on. [MENTION=23751]Maxperson[/MENTION] has no problem ignoring the fact that 5e flat out stats that NPC's that aren't a threat don't need stats, why should he bother actually reading this either?

 

[extralead]

For 5e, I am using 1st-Edition Method III -- http://www.mithrilandmages.com/utilities/1ECharacterAttributes.php -- (select the drop down to get to roman-numeral III before Generate).

When using this method, the DM must force the players to create all aspects of their character including Faction, Background, Race, Class, and potential subclass before rolling Method III. Equipment will likely be the only process after rolling for them.

If racial bonuses or feats push the attribute past 17, in some campaigns it may be best to leave the maximum stat at 17 after these adjustments, even if an 18 is rolled. In other situations, these can start at 20 with those racial and/or feat bonuses. Like, really, who cares? I like the idea that a Half-Orc Sorcerer might start with a 9 Charisma but also have a 17 Strength and a 20 Constitution. Oh well. Play hard.

 

[Oofta]

It's cookie-cutter NPCs I'm after in this case.

The specifications I'd determined by measuring and statictically analysing a whole lot of trees that I could access, to model those I could not access.

Except we can measure it for one stat, more or less, using real-world IQ scores or equivalent; and we can thus know pretty much exactly what the distribution should be across a population. It's not a big stretch to then for game purposes apply that same distribution to the other five stats.

We agree that 3d6 gives too loose of a bell curve for any stat.

I think we agree that the distribution for any stat across a population follows a bell curve of some sort, though we could debate for years how to arrive at a means of randomly generating individual scores along that curve.

I hope we agree that those six bell curves (one for each stat) won't all nicely cancel out for every given individual; that some individuals will be genetically luckier than others.

Where we disagree is how any of this affects PC generation as opposed to modelling the population at large.

With me so far? Or, if not, where am I going wrong?

Lanefan

There was a long argument the last time this came up. Suffice to say I agree - that as flawed as IQ is, it's the best we've got. The range matches up rather well IMHO, high is around 180-200, average is a little over 100, minimum is around 30. So just take IQ and divide by 10, right?

Use something like:


To get something like


I would be OK with that. Match up the distributions, roll a percentile dice for your average Joe or Jane.

For reasons I still don't quite get, this was a controversial opinion. IIRC correctly it had to do with standard deviations, etc. etc. To be honest I just got tired of arguing about it.

One issue is that it's generally agreed nowadays that IQs below 55-ish and above 155-ish start to get kind of meaningless. Some people estimate Einstein's IQ at 185 for example (seems about right) but others ... well let's just say that IQ scores are controversial.

So if you fudge a little bit with the standard deviation breakdowns to convert to 3-18 scale, I'd be OK with .1% having a 3 or an 18. Roll a D1000 and get a 1000? You have an 18 int.

The issue would be the numbers between no bonus or penalty and that +/- 4.

I don't think that adds any real value to the game, but if I had to generate ability scores for my NPCs, this would be my starting point.

 

[Tony Vargas]

Variability.

People in the real world are not "balanced" anywhere near as evenly as we would be if we were all using the same stat array. The range between best and worst simply isn't big enough, and we don't all perfectly trade off at being just as good at one thing and just as bad at another.

Adventurers aren't the whole population, though. People in the real world who have gone through similar selection processes, OTOH, likely do have more similar capabilities. Professional athletes or elite soldiers or whatever, would be 'realistically' analogous to adventurers. Regardless of the character generation method you choose, the DM has complete latitude with the rest of the population, and thus, variability is exactly what he decides to make it.

And, besides point-buy doesn't give everyone the same stats or even the same array (there's a lot of possible different arrays under point buy, nor am I aware of a consensus 'optimal' point-buy array), there'll like as not be 6 different arrays in a party of 6 created via point-buy - for that matter, random generation, being random, could give two or more characters in the party the same array.

But, you are right about one thing: /balance/ is not a feature built into reality the way it needs to be in games to make them playable. Because nobody can just not show up to a session of reality without even phoning ahead. ;P

Seriously, though, (as seriously as possible, which is not very) balance (if you like that sorta thing, and if you do, why are you playing D&D) is a strength of point-buy or standard array and/or a weakness of random generation. 'Realism' isn't simply 'not balance' - something can be unrealistic /and/ imbalanced, for instance - but, balance is often going to get in the way of realism, especially for those instances in which reality kinda sucks.

It's cookie-cutter NPCs I'm after in this case.

That's entirely up to the DM: they're his cookies, he can cut them how he likes.

(Or did you mean NPCs who cut cookies - like the 'Keebler' elven sub-race?)

 

[Lanefan]

There was a long argument the last time this came up. Suffice to say I agree - that as flawed as IQ is, it's the best we've got. The range matches up rather well IMHO, high is around 180-200, average is a little over 100, minimum is around 30. So just take IQ and divide by 10, right?

Use something like:
View attachment 90247

Close enough.

To get something like
View attachment 90246

This doesn't sell me on linear bonuses. 0-1-2e had it right, where the bonuses/penalties j-curved as you got closer to one extreme end or the other.

I would be OK with that. Match up the distributions, roll a percentile dice for your average Joe or Jane.

So if you fudge a little bit with the standard deviation breakdowns to convert to 3-18 scale, I'd be OK with .1% having a 3 or an 18. Roll a D1000 and get a 1000? You have an 18 int.

Sure. This almost gets granular enough to be useful as a representative model.

The issue would be the numbers between no bonus or penalty and that +/- 4.

See above re linear bonuses.

Lan-"now wondering if generating PCs on that d1000 system with advantage on each roll would work, and what it'd give for stats"-efan

 

[Hriston]

Variability.

People in the real world are not "balanced" anywhere near as evenly as we would be if we were all using the same stat array. The range between best and worst simply isn't big enough, and we don't all perfectly trade off at being just as good at one thing and just as bad at another.

Ditto for if we were all built by point-buy - there'd be a bit more variability but nowhere near enough range. Take intelligence by itself: people exist with IQs of 30, and of 180 or more, across a bell curve that peaks somewhere in the 110 range I think. Now there's no denying that this real-world bell curve is a bit tighter than rolling 3d6 produces, but there's also no denying its existence and there's also no denying that the 3d6 model is a reasonably elegant, if imperfect, attempt to model it.

The game tries to model that same bell curve across six stats instead of just one; in all cases the resulting bell curve is too loose, but in all cases the model sort of works as intended.

I think you missed the point of my question. I wasn't asking why rolling is more realistic than everyone taking the same array or everyone using point-buy. My question was why it's more realistic for an individual character to have a set of rolled scores than it is for the character to have the same scores derived from another method.

Let's say there are three PCs in a group that uses 4d6 drop lowest. The player of the first PC rolls 15, 14, 13, 12, 10, 8, which just so happens to be close to the most likely outcome of such a roll. The player of the second PC rolls 14, 14, 14, 12, 9, 9. And the third player rolls 10, 10, 10, 10, 10, 10 for his character's abilities.

Now let's say there's another group with three PCs, but this group likes to use a variety of methods to set its scores, but they never roll dice. One player uses the standard array for her character. Another player uses point-buy to build an array of 14, 14, 14, 12, 9, 9. The last player wants to play a commoner masquerading as a hero and uses the scores of the Commoner NPC for his character.

Why is the first group more realistic?