D&D General Explain Bounded Accuracy to Me (As if I Was Five)

[Jahydin]

My friend's one set of houserules works on a +8 proficiency scale, +4 ability max, and a d12 instead of a d20.

I like it, personally, but wouldn't mind just bumping it to +10 proficiency, +5 ability, and d20. Expertise would be a +50% on proficiency. IMO the best "bounds" are when "maximum skill" equals the maximum of the die rolled.

Now, 5E is pretty close to that anyway, +17 "max" with expertise and +5 ability, but I agree I don't like that max proficiency at +6 is only one better than "max" ability at +5.

Really cool!

In my last campaign I just used double Prof bonus and that helped. Can't remember what I did with Expertise...? Advantage maybe? Also used 2d10 for Skill checks, d20 for everything else (combat being more random makes sense to me!).

Since then, I realized RAW is fine as long as you use skill checks:

• When it actually matters for in-game reasons.
• Use Advantage and Disadvantage liberally to bump things in sensical directions.
• Know it's okay to not roll and just make a call when the desired task is trivial/impossible. (This should happen a lot.)
  • Take into consideration Class/Background/Proficiency when you do.

But I do tweak skills just a bit to make them more impactful:

• Handle Animal proficiency gives you a tiny animal companion.
• Investigation is used for searching, not Perception.
• Medicine lets you identify poisons and determine cause of death.
• Etc.

 

[Crimson Longinus]

How random should the skill rolls and such be is an interesting debate. It is true that 5e is very swingy; more so than I think it is realistic. However, I also understand why it is so. In my experience people seem to like this randomness. Everyone has a chance, and even an expert usually has the excitement of the die roll mattering. Guaranteed successes and failures are actually not that fun. I feel games where the character competence has more impact should be generally constructed so that there is usually a degree of success or failure that matters. Then it is still exiting to roll, as there is still uncertainty about the outcome even if you knew you probably get some sort of a success. This is harder to design well than a simple pass or fail system though.

 

[EzekielRaiden]

No I don't. Sentiment was expressed that it would be somehow silly or boring for high level characters to be threatened by something as feeble as goblins. But it works fine, and it is not an issue in either a game or a novel.

If the point was "this was done in a novel, thus it is fine to do in a game" and then someone points out all the ways that there are things that are fine to do in novels that would be terrible to do in a game...you have not established that the thing is fine to do in a game.

Note, I am not saying it isn't fine to do this in a game. I think it is. But the thing you wanted to establish simply hasn't been, because the claimed reason--"if it's fine for a novel, it's fine for a campaign"--is patently false.

In short bounded accuracy in 5e fails at pretty much everything it promised and creates a bunch of damned if you do damned if you don't type problems for the gm to be blamed for either way

I mean, it's quite cold comfort, but I kinda predicted...pretty much all of this during the D&D Next playtest. Design isn't just some pure ineffable expression of auteur intent beyond any possibility of analysis or technical criticism. Game design is a technology. It can be used for better or worse ends, and it can be evaluated as to whether it achieves the ends for which it was designed.

As a general rule, I think it is reasonable to ask that successful products should care about doing the things they claim to do.

Why is giving an option a problem?

Depends on what the option is, and how the giving is done. I could use some rather....pointed examples, but I suspect that would be excessively inflammatory. Instead, we can just construct simple (and intentionally hyperbolic) examples of "new options" for D&D content. Imagine if we added a 1st-level spell that did 100d6 "chaos" damage (meaning, a damage type which nothing in 5e has any immunity nor resistance to) to all enemies (not allies) in a 100' radius, no saving throw, upcasting for a further 100d6 per spell level. That's merely adding an option! But it's quite clearly a problem as well, or at least I should hope you would consider it so, given how hyperbolically bad I've constructed it to be. We could do the same thing for races (Kryptonian, +10 to all ability scores and the ability to cast several offensive and defensive spells at-will), classes (make a new class that has all the class features of Druid, Cleric, and Wizard), backgrounds (you are proficient in all skills; pick two skills that gain expertise), you name it.

TL;DR: "Giving an option" that dwarfs or distorts the value of all other options by comparison can, in fact, be a pretty big problem. The burden, of course, is to show that a particular option does distort things in this way. But I don't think it's hard to argue that giving players the option to never do anything particularly dangerous or unsafe, but instead stick with the exact same opponents in greater numbers (which...generally cannot be brought to bear because terrain effects, area denial, and action economy do not permit effective use of massive numbers of combatants) will result in them choosing to do that consistently, rather than actually facing new and interesting, but more risky, dangers instead.

TTRPG players are usually cautious to a fault, unless they're full-on murderhobos...who wouldn't care either way, so it's not like their opinion is changing the results any.

The thing is, even with bounded accuracy, you don't need to put goblins against high level characters. But if I want, I can. That I can do that in no way negatively affects you.

Except that it does. Because this is not "option A is intense tactical combat with a need to continually face tougher opponents, while option B is intense strategic combat against teeming hordes." It is replacing option A with option B. Which was the whole point--get rid of the feeling of progress from becoming an insurmountable obstacle for your old enemies, and instead (attempt to) provide a feeling of progress from facing more enemies of the same kind.

You will probably not be surprised to know that I, and many others, find this unsatisfying.

 

[EzekielRaiden]

How random should the skill rolls and such be is an interesting debate. It is true that 5e is very swingy; more so than I think it is realistic. However, I also understand why it is so. In my experience people seem to like this randomness. Everyone has a chance, and even an expert usually has the excitement of the die roll mattering.

Surprising. It's quite rare for a 5e fan to actually take any issue at all with how swingy it is. Most either deny that it is swingy in the first place, or defend it (often the former becomes the latter, I find.)

My main problem is, I genuinely don't think everyone should have a chance, on the one hand, and yet on the other hand I very much think an expert should truly outgrow some challenges, becoming not merely quite likely but guaranteed to succeed. That is, in many ways, how we judge degrees of competence at a task, after all--how much a person's success depends on making the right choices/taking the right actions, rather than on luck or happenstance. A climber who fails 20% of perfectly ordinary climbing attempts is someone I would consider not very good at climbing, especially if he's supposed to be an "expert."

Guaranteed successes and failures are actually not that fun.

For you. (A message you've been quite keen to make to me in the past, I'll note.)

I feel games where the character competence has more impact should be generally constructed so that there is usually a degree of success or failure that matters. Then it is still exiting to roll, as there is still uncertainty about the outcome even if you knew you probably get some sort of a success. This is harder to design well than a simple pass or fail system though.

Oh, certainly, degrees of success would be lovely. If only we had some kind of system that could judge group-level degrees of success, say, by succeeding or failing at various checks along the route toward a final goal. Or if we had, say, a page or two in the DMG going over what typical tasks are like, so we could get a sense for what near-success or near-failure is, and thus graduate from e.g. costly failure to general failure to almost-successful to barely-passing to passing comfortably to passing with flying colors. Those tools would be incredibly useful, but alas, they're definitely totally impossible and no edition of D&D has ever even attempted such an unachievable dream.

 

[Crimson Longinus]

If the point was "this was done in a novel, thus it is fine to do in a game" and then someone points out all the ways that there are things that are fine to do in novels that would be terrible to do in a game...you have not established that the thing is fine to do in a game.

Note, I am not saying it isn't fine to do this in a game. I think it is. But the thing you wanted to establish simply hasn't been, because the claimed reason--"if it's fine for a novel, it's fine for a campaign"--is patently false.

You pointed ways which games are different from novels which were irrelevant to the point. Thematically it is dine for powerful characters to be threatened by a bunch of weak foes.

Depends on what the option is, and how the giving is done. I could use some rather....pointed examples, but I suspect that would be excessively inflammatory. Instead, we can just construct simple (and intentionally hyperbolic) examples of "new options" for D&D content. Imagine if we added a 1st-level spell that did 100d6 "chaos" damage (meaning, a damage type which nothing in 5e has any immunity nor resistance to) to all enemies (not allies) in a 100' radius, no saving throw, upcasting for a further 100d6 per spell level. That's merely adding an option! But it's quite clearly a problem as well, or at least I should hope you would consider it so, given how hyperbolically bad I've constructed it to be. We could do the same thing for races (Kryptonian, +10 to all ability scores and the ability to cast several offensive and defensive spells at-will), classes (make a new class that has all the class features of Druid, Cleric, and Wizard), backgrounds (you are proficient in all skills; pick two skills that gain expertise), you name it.

This is an absurd example, and not at all comparable at all to what we are actually discussing.

TL;DR: "Giving an option" that dwarfs or distorts the value of all other options by comparison can, in fact, be a pretty big problem. The burden, of course, is to show that a particular option does distort things in this way. But I don't think it's hard to argue that giving players the option to never do anything particularly dangerous or unsafe, but instead stick with the exact same opponents in greater numbers (which...generally cannot be brought to bear because terrain effects, area denial, and action economy do not permit effective use of massive numbers of combatants) will result in them choosing to do that consistently, rather than actually facing new and interesting, but more risky, dangers instead.

That you can sometimes use lower level foes doesn't mean you always must. High level foes do in fact exist in 5e.

Except that it does. Because this is not "option A is intense tactical combat with a need to continually face tougher opponents, while option B is intense strategic combat against teeming hordes." It is replacing option A with option B. Which was the whole point--get rid of the feeling of progress from becoming an insurmountable obstacle for your old enemies, and instead (attempt to) provide a feeling of progress from facing more enemies of the same kind.

You will probably not be surprised to know that I, and many others, find this unsatisfying.

Just no. You now have option to use A or B as the situation warrants. I don't really understand why I need to keep repeating this. That high level characters can fight goblins doesn't mean that this is all they do or even that it must ever be done if you don't want to.

 

[Crimson Longinus]

Surprising. It's quite rare for a 5e fan to actually take any issue at all with how swingy it is. Most either deny that it is swingy in the first place, or defend it (often the former becomes the latter, I find.)

Oh, it definitely is swingy. I think it is fine game, but I am not 5e fan in a sense that I think it is best game ever. But I am just trying to point out that there is a reason why it is like it is, and whether you like it is more of a preference thing.

My main problem is, I genuinely don't think everyone should have a chance, on the one hand, and yet on the other hand I very much think an expert should truly outgrow some challenges, becoming not merely quite likely but guaranteed to succeed. That is, in many ways, how we judge degrees of competence at a task, after all--how much a person's success depends on making the right choices/taking the right actions, rather than on luck or happenstance. A climber who fails 20% of perfectly ordinary climbing attempts is someone I would consider not very good at climbing, especially if he's supposed to be an "expert."

My intuition of "what makes sense" broadly agrees with you. And of course even with swingy 5e, some tasks become automatic, and some lie beyond capabilities of the characters. I really disliked when they tried to introduce the autofail on one and autosuccess on 20 for skill checks in one of the playtests. But realistically skill should probably matter more, but I am not convinced it would necessarily lead to better play experience for a group based action adventure game 5e is. Basically what the designers want that in a random group of character they have some chance to succeed in most tasks even if an expert doesn't happen to be present.

For you. (A message you've been quite keen to make to me in the past, I'll note.)

Yes. It is also what I have observed to be common attitude over the decades I've played RPGs. But of course that is still anecdotal, and it is hard to say what "most people" actually prefer.

Oh, certainly, degrees of success would be lovely. If only we had some kind of system that could judge group-level degrees of success, say, by succeeding or failing at various checks along the route toward a final goal.

There are group checks.

Or if we had, say, a page or two in the DMG going over what typical tasks are like, so we could get a sense for what near-success or near-failure is, and thus graduate from e.g. costly failure to general failure to almost-successful to barely-passing to passing comfortably to passing with flying colors. Those tools would be incredibly useful, but alas, they're definitely totally impossible and no edition of D&D has ever even attempted such an unachievable dream.

Yes, DMG is bad and it should have these things. Nothing in 5e's basic architecture prevents this.

 

[EzekielRaiden]

You pointed ways which games are different from novels which were irrelevant to the point. Thematically it is dine for powerful characters to be threatened by a bunch of weak foes.

But not because it's fine in novels. That is precisely the point. Your reasoning does not defend the point made.

This is an absurd example, and not at all comparable at all to what we are actually discussing.

The claim was that it is never a problem, no matter what, to add an option; that doing so could not ever be an issue. This claim is false. The hyperbolic examples simply demonstrate a situation that no one could possibly accept.

Hence: Simply arguing "we're just adding an option, it's not like that could possibly hurt you" is false. You must further demonstrate that it is an appropriate, fitting option amongst the ones already in play. Which is precisely the thing others are complaining about. That this option drives out other alternatives.

That you can sometimes use lower level foes doesn't mean you always must. High level foes do in fact exist in 5e.

Except that they, too, have been subjected to the mathematical flattening! You can't have it both ways. You can't have flattened math and not-flattened math. It must be one or the other--by definition. If you have un-flattened the math for some options, then by definition the math in general is not flattened!

Just no. You now have option to use A or B as the situation warrants. I don't really understand why I need to keep repeating this. That high level characters can fight goblins doesn't mean that this is all they do or even that it must ever be done if you don't want to.

Just no. You can only fight flattened-math opponents. Those are the only ones allowed to exist in 5e. Yes, some of them will be at the upper end of what little numeric progression is permitted.

The highest possible AC for a creature in 5e is 25. The lowest is, naturally, 0. Even accounting for the difference in maximum level, reducing it to ~34, that's 9 points lower than the maximal 4e equivalent (which, perhaps ironically, is Tiamat vs Bahamut.) Even if we look at a more typical top end, it's still ~32 vs ~22. Meaning the absolute strongest, toughest opponents a player would typically face...have approximately the same defenses as a highly-defensively-built PC. (Plate + shield + defense fighting style = AC 21. Something a character can achieve at, roughly, level 6, depending on how quickly they get the money required to buy a suit of plate.)

My intuition of "what makes sense" broadly agrees with you. And of course even with swingy 5e, some tasks become automatic, and some lie beyond capabilities of the characters. I really disliked when they tried to introduce the autofail on one and autosuccess on 20 for skill checks in one of the playtests. But realistically skill should probably matter more, but I am not convinced it would necessarily lead to better play experience for a group based action adventure game 5e is. Basically what the designers want that in a random group of character they have some chance to succeed in most tasks even if an expert doesn't happen to be present.

But "some chance to succeed" is mostly pointless if...well, "60% of the time, it fails every time," to twist a popular meme.

There are group checks.

Group checks do not even slightly do what I described.

Yes, DMG is bad and it should have these things. Nothing in 5e's basic architecture prevents this.

Well, other than an active and pervasive effort to avoid their inclusion and a culture of play actively hostile to the concept of statistical testing of mechanics.

 

[Crimson Longinus]

But not because it's fine in novels. That is precisely the point. Your reasoning does not defend the point made.

It is fine in games and novels because it is thematically fine. In fact any narrative from LotR would be fine in a game. Your examples of what would not be fine were conjectures about behind the scenes decisions that led to those narrative, but those were just your invention, and we could have arrived to the narrative in other means.

The claim was that it is never a problem, no matter what, to add an option; that doing so could not ever be an issue.

Was it?

This claim is false. The hyperbolic examples simply demonstrate a situation that no one could possibly accept.

Hence: Simply arguing "we're just adding an option, it's not like that could possibly hurt you" is false. You must further demonstrate that it is an appropriate, fitting option amongst the ones already in play. Which is precisely the thing others are complaining about. That this option drives out other alternatives.

Yes. But adding any option is fine was never at least my argument. I certainly have opposed adding some options in the past. My argument that adding this option is fine.

Except that they, too, have been subjected to the mathematical flattening! You can't have it both ways. You can't have flattened math and not-flattened math. It must be one or the other--by definition. If you have un-flattened the math for some options, then by definition the math in general is not flattened!


Just no. You can only fight flattened-math opponents. Those are the only ones allowed to exist in 5e. Yes, some of them will be at the upper end of what little numeric progression is permitted.

The highest possible AC for a creature in 5e is 25. The lowest is, naturally, 0. Even accounting for the difference in maximum level, reducing it to ~34, that's 9 points lower than the maximal 4e equivalent (which, perhaps ironically, is Tiamat vs Bahamut.) Even if we look at a more typical top end, it's still ~32 vs ~22. Meaning the absolute strongest, toughest opponents a player would typically face...have approximately the same defenses as a highly-defensively-built PC. (Plate + shield + defense fighting style = AC 21. Something a character can achieve at, roughly, level 6, depending on how quickly they get the money required to buy a suit of plate.)

So what? In 4e you never actually fought enemies wildly out of your level range. The instruction was to use roughly level appropriate opponents so that you would not never face those extreme numbers, that would break the game. The game even had awkward kludges like minions to deal with the math failure. 5e approach where the math just works and you can fight enemies of much lower and higher level without having to change the their stats is clearly superior. But in practice the end result is the same: you face the enemies whose stats are at such range that you can hit them and they can hit you.

But "some chance to succeed" is mostly pointless if...well, "60% of the time, it fails every time," to twist a popular meme.

I don't understand what you're trying to say here. It is different thing to get to at least roll than just being told flat "no."

Group checks do not even slightly do what I described.

They definitely "even slightly" do.

Well, other than an active and pervasive effort to avoid their inclusion and a culture of play actively hostile to the concept of statistical testing of mechanics.

If you say so.

 

[ezo]

Surprising. It's quite rare for a 5e fan to actually take any issue at all with how swingy it is. Most either deny that it is swingy in the first place, or defend it (often the former becomes the latter, I find.)

I don't think it is really surprising at all. I've read many posters (as fans) complain about the swinginess of the d20 mechanic, especially now that your "typical" (across levels) bonus is about +8, and the "typical" maximum difference between 1 and 20 is just +6 (from +5 to +11).

What I always found interesting is that success in most situations for a competent PC is about 65% (8 or higher on the d20), which means 35% of the time they "do not make progress" (not necessarily "failure"). So, while not experts, it seems like failing roughly 40% of the time is acceptable, perhaps even preferred.

Now, I've read about people suggesting 2d10, 3d6, 3d20 take middle, and probably others. And if you look at these the probably of 8 or better is 0.790, 0.838, and 0.718, respectively, decreasing the chance of "not making progress" to about 20-30%. To many people, that seems better.

Ok, but let's bump up the AC/DC for the PC. Instead of 8 needed on the d20, how about 13 and 18 even?

At 13, d20 is 40% success. 2d10, 3d6, and 3d20tM are about 36%, 26%, and 35%, respectively.
At 18, d20 is 15%, while the others are 6%, 0.5%, 6%.

So, while the other methods remove the swinginess for the d20 and cluster most results around the centers, they also make the truly difficult tasks so hard that it almost becomes an exercise in futility. IMO I don't think many people really want that, either.

As I see it, you either live with the swinginess, or accept that common tasks will be fairly easy, but those really high AC/DCs will be almost impossible. Personally, I find the latter of the two more acceptable, but I can understand why others prefer the d20.

 

[MuhVerisimilitude]

I don't think it is really surprising at all. I've read many posters (as fans) complain about the swinginess of the d20 mechanic, especially now that your "typical" (across levels) bonus is about +8, and the "typical" maximum difference between 1 and 20 is just +6 (from +5 to +11).

What I always found interesting is that success in most situations for a competent PC is about 65% (8 or higher on the d20), which means 35% of the time they "do not make progress" (not necessarily "failure"). So, while not experts, it seems like failing roughly 40% of the time is acceptable, perhaps even preferred.

Now, I've read about people suggesting 2d10, 3d6, 3d20 take middle, and probably others. And if you look at these the probably of 8 or better is 0.790, 0.838, and 0.718, respectively, decreasing the chance of "not making progress" to about 20-30%. To many people, that seems better.

Ok, but let's bump up the AC/DC for the PC. Instead of 8 needed on the d20, how about 13 and 18 even?

At 13, d20 is 40% success. 2d10, 3d6, and 3d20tM are about 36%, 26%, and 35%, respectively.
At 18, d20 is 15%, while the others are 6%, 0.5%, 6%.

So, while the other methods remove the swinginess for the d20 and cluster most results around the centers, they also make the truly difficult tasks so hard that it almost becomes an exercise in futility. IMO I don't think many people really want that, either.

As I see it, you either live with the swinginess, or accept that common tasks will be fairly easy, but those really high AC/DCs will be almost impossible. Personally, I find the latter of the two more acceptable, but I can understand why others prefer the d20.

That's a solvable problem. The more dice you add the more stable the result and the less random the outcome is.

It's frankly baffling to me that people are arguing that there shouldn't be any ACs too hard to hit for a low level character. Why? Consider something like, say, flight. The system does not at all care about creatures like dragons having flight, even though given the right circumstances flight is equivalent to infinite AC.

It's not like a character is just passively standing there letting their AC soak attacks either. AC is an active defence that involves dodging incoming attacks and moving around. Why is it so important that even high level characters should be hittable?