D&D 5E Point Buy vs Rolling for Stats

[Lanefan]

Ahh, so we're back to mechanics telling us what our game worlds should look like. No thanks. I don't play that way. My game world looks how I want it to look and I use the mechanics during the game to resolve different contests, nothing more.

Well if the mechanics don't tell us what the game world looks like then the game world must be driving the mechanics, since they somehow have to agree. Either that or all hope of internal consistency goes wandering down to the pub for a beer, which...while never a bad idea...isn't really a desired result.

And, how can it not include the best and worst stat? Are you saying that die rolling should be constrained somehow so that those best and worst scores are not possible? IOW, applying the 8-15 array? Otherwise, we're back to that 20% best/worst ratio. What exactly are the stats for best and worst if it's not 3 and 18?

Easy. 18 is the highest you can roll; but (in 5e) 20 is the highest you can get to. 3 is the lowest you can roll; (in 5e) 1 is the lowest you can be. Artificially limited extremes, if you ask me, but there it is.

Other editions were more open-ended - you'd always start by rolling in the 3-18 range but your stats could end up who-knew-where given enough time and-or luck and-or wishes and-or whatever, and particularly in 3e there was no hard-coded "best" or "worst" even for commoners.

My prefernece is that the stats be open-ended but that the true extremes (i.e. anything outside 3-18) be increasingly harder - eventually approaching but never quite reaching impossible - to achieve.

Lanefan

 

[Maxperson]

"Utterly realistic"? Seriously? That's so far away from anything approaching realistic that straight 10's for all NPC's looks quite reasonable. 1 in 5? That's not realism. That's not even close to believable for me. Never minding that you're insisting that 5e NPC's use 4d6-L which is even further from anything believable.

See, there you go again not understanding realism. Quite literally everything approaches realistic since everything greater than some unknowable chaos void(no realism at all, which doesn't exist) is on the realism scale somewhere. Even magic and dragons have some realism to them. Both exist as real life myths/legends.

My point is, even if die rolling was more realistic than point buy or array (which I'm not particularly convinced it is), any gain in realism is dwarfed by the degree of unbelievability that die rolling creates.

So what.

IOW, adding a drop of red to a gallon bucket of blue paint does not make it purple. Is it more purple than a bucket of blue paint? Sure, maybe. But, who cares? It's still blue. Any purple is completely washed way by the blue.

It's not that slight, though. The amount of red(realism) that rolling adds may not make the bucket purple, but it's enough to slightly alter the shade of blue in a noticeable way.

But, if you're seriously arguing that die rolling somehow objectively results in a more realistic world, then we are just not going to agree. Die rolling creates an even more unbelievable world than straight 10's. At least with straight 10's for NPC's, everyone is average. You don't have 1 in 5 NPC's that are off the charts.

So you'd rather believe that it's more realistic that 99.9% of the world are statistical clones, than for them to have variety like the real world, and random stats like the real world.

And, let's be honest, no DM will EVER stick to a die roll for an NPC if he or she is unhappy with the result.

If the DM needs an NPC to have a a stat with a specific number, he's not going to roll it. You're talking apples and oranges here.

 

[Maxperson]

And, how can it not include the best and worst stat? Are you saying that die rolling should be constrained somehow so that those best and worst scores are not possible? IOW, applying the 8-15 array? Otherwise, we're back to that 20% best/worst ratio. What exactly are the stats for best and worst if it's not 3 and 18?

The best is 30(You can get over 20 with boons and items) and the worst is 1. None of the stat methods include the best or worst that exists in the world.

 

[Tony Vargas]

The best is 30(You can get over 20 with boons and items) and the worst is 1. None of the stat methods include the best or worst that exists in the world.

So a range of 8-15 is not so meaningfully different from 3-18, since neither are "the full range."

The stats resulting from die rolling are undeniably going to be more realistic than those resulting from either point buy or standard array.

That can't possibly be true, as they can give the exact same results.

I think what you mean is that the result set of generating thousands of arrays via a random method is going to look more like a random sampling of a population - if you had a way to measure those 6 broad/fuzzy attributes on a uniform 1-20 scale and if you only looked at the numbers, not their distribution among said 6 attributes (realistically, the comparatively harsh life of the majority of a medieval population is going to skew towards, for instance, reasonable CON, as disastrously-low-CON individuals will likely die quite young).

If you're only looking at the results of one character or one group of a half dozen or so, the difference in realism-by-plausible-distribution-of-attributes between point-buy and random is going to be trivial - even standard array might not stand out too much in such a small sample size.

The 'realism' side is more theoretical or hypothetical: if you were to use a random method to generate stats for every NPC/monster/whatever in the setting, random - using different random methods for different populations, of course, you wouldn't want to roll squirrel STR on 4d6k3 - would presumably produce a more naturalistic-seeming distribution than array (using different arrays for different populations for the same reason) or point-buy (using different point totals and limits, naturally)

I have no problem with someone saying the like die rolling because it feels more realistic or believable to them. No problem.

I do have a problem with flat statements claiming that die rolling is more realistic.

'Realism' has been a fraught issue virtually since the game's inception. OT1H, realism in an objective sense is actually easy to test for: Does the game simulate reality like a scientific model? Can you run the same scenario with the same starting conditions & decisions in D&D, and, independently, in reality, and get similar results? Obviously not. D&D is not remotely realistic in the way a simulation or theory would be an accurate model.

OTOH, a subjective feeling of realism, or immersion, or verisimilitude can, being subjective vary wildly from one person to another, and, thus makes a lovely stalking horse for any other preference that maybe might not seem as sympathetic. Like really wanting to play that multi-18 ubercharacter, or really wanting to CharOp that 'realistic concept' to the nth degree.

So, ultimately, it's worse than useless as a quality by which a game, or even chargen method, can be assessed.

Well if the mechanics don't tell us what the game world looks like then the game world must be driving the mechanics, since they somehow have to agree.

There's a lot of room for imagination to make up the inevitable gap between the two.

 

[Caliban]

The stats resulting from die rolling are undeniably going to be more realistic than those resulting from either point buy or standard array.

After 100+ pages of this exact thing being denied, that's the best you have? Bwhahahaha...

 

[Lanefan]

So a range of 8-15 is not so meaningfully different from 3-18, since neither are "the full range."

That can't possibly be true, as they can give the exact same results.

For a single character, yes. For a hundred characters, no.

I think what you mean is that the result set of generating thousands of arrays via a random method is going to look more like a random sampling of a population - if you had a way to measure those 6 broad/fuzzy attributes on a uniform 1-20 scale and if you only looked at the numbers, not their distribution among said 6 attributes (realistically, the comparatively harsh life of the majority of a medieval population is going to skew towards, for instance, reasonable CON, as disastrously-low-CON individuals will likely die quite young).

Good point re Con; I suspect low Wisdom would also self-cull to some extent.

That said, if the char-gen system can at least vaguely reflect the system used to generate the overall population it becomes far easier to picture the characters as having been drawn from that overall population, and having been part of it during their lives so far.

'Realism' has been a fraught issue virtually since the game's inception. OT1H, realism in an objective sense is actually easy to test for: Does the game simulate reality like a scientific model? Can you run the same scenario with the same starting conditions & decisions in D&D, and, independently, in reality, and get similar results? Obviously not. D&D is not remotely realistic in the way a simulation or theory would be an accurate model.

When fantasy elements such as magic are stripped away from the game world, ideally the results would be somewhat similar; giving a solid and believable foundation on which to place the fantasy elements. And where the results are not similar there would be a clear and reasonable (or at least rational) explanation why; this explanation might be as simple as the DM deciding she wants something to be different in her world (e.g. she wants a clear sky to be orange instead of blue and explains it as caused by a different chemical composition of either the planet's atmosphere or the planet's main light source).

After that, all that's needed is that the game world remain consistent with itself. PC generation methods being in isolation from NPC generation methods violates that consistency.

After 100+ pages of this exact thing being denied, that's the best you have?

Doesn't matter if the denial is on page 1 or page 100, it's still flat wrong and needs to be challenged.

Lanefan

 

[Hussar]

After that, all that's needed is that the game world remain consistent with itself. PC generation methods being in isolation from NPC generation methods violates that consistency.

LOL. And yet, they have been isolated from each other since day 1 without any real problem. And remain so today since no PC get's to choose his or her stats, yet, I can, by the advice in the DMG, choose the stats for every single NPC, or, indeed have NPC's without any stats at all.

Again, it all comes down to the idea that the mechanics MUST apply to everything in the world. That simply isn't, and never has been, true. You can choose to do it that way. But, it isn't required.

 

[Caliban]

Doesn't matter if the denial is on page 1 or page 100, it's still flat wrong and needs to be challenged.

Lanefan

Except it's not "undeniable". What is undeniable is that is has been in fact been denied!

Making such an arrogant claim in a forum full of geeks is just asking for it.

It's your opinion that rolling is more realistic, partially as justification of your preference for rolling stats. It feels more realistic to you. And that's fine.

But being pissy about other people not sharing the same opinion about a minor aspect of a game is...eh, it's what gamers do, I guess. But don't expect it to go unchallenged.

 

[Tony Vargas]

For a single character, yes. For a hundred characters, no.

I was thinking thousands, but yes, that's the point. You don't roll up thousands, or even hundreds, of PCs in a typical campaign (maybe dozens in a really Killer one), let alone keep statistics about their, well, statistics.

So they hypothetical PC bell curve vs the hypothetical population bell curve is moot.

Good point re Con; I suspect low Wisdom would also self-cull to some extent.

That said, if the char-gen system can at least vaguely reflect the system used to generate the overall population it becomes far easier to picture the characters as having been drawn from that overall population, and having been part of it during their lives so far.

Remember, there /is no system used to generate the overall population/. The overall population isn't generated - it's more like a painted back-drop in a play, bits implied to be part of it step on stage as they're encountered - and they're likely generated by DM whim/judgement...
...so if you want chargen to reflect that...

When fantasy elements such as magic are stripped away from the game world...

Fantasy elements are only a bar against RL-realism. You could check D&D against a fantasy source, too. Does a D&D campaign starting with the characters and situations at the end of the Hobbit output LotR? No. Never, not even if you try to force it. Objectively it fails as a simulation.

After that, all that's needed is that the game world remain consistent with itself. PC generation methods being in isolation from NPC generation methods violates that consistency.

Internal consistency is an easier and more reasonable bar than objective realism. But, it's not something that matters to the choice of generation method(s). If the DM chooses to use 4d6k3 for PCs, it would be consistent enough for him to use 3d6 when he generates 'ordinary' NPCs, at all, and 4d6k3 for 'special'/important ones meant to be the equal of PCs - and possibly something more generous for NPCs meant to be superior. If he chooses to use Point Buy for PCs, he can use point buy for those NPCs he assigns stats to at all - lower point-buy with tighter limits for ordinary NPCs, PC-equivalent for exceptional ones, etc...

 

[Hriston]

As long as your dice rolls are independent and identically distributed (which we always assume, even if it's not precisely correct), xdn will look more and more normal-ish as x gets larger. In fact, in a (very informal) sense, that's what the normal distribution is* and why it is interesting / useful.

* xdn is a special case; it's actually much more general. Google 'central limit theorem' if interested; the proof is kind of math-y, but the statement is really pretty straightforward.

Eureka! (15d2)-12 conforms more closely to a normal distribution than any other dice-rolling method yet, and it covers the 3-18 range as well! Interestingly, it comes within 1-2% of the frequency of all results of 3d{2,3,3,4,4,5}, except for 10 and 11, which are slightly less likely using this method. I'd say the averaging method is a good approximation, but if the remote chance of rolling 3 or 18 is important, this method produces a distribution that's very close to what you'd expect to find in the real world.