One of the most talked about changes in the One D&D playtest is the decision to make all natural 1’s auto-failures and all natural 20’s auto-successes.

My first gut reaction to this was: That’s a terrible idea!

Upon further reflection, however, I’ve realized that this reaction is primarily built on my experience with pre-5^th Edition versions of D&D, and that under the design principles of 5^th Edition it’s probably irrelevant. (We’ll come back to that.)

But here’s what I don’t get:

You’re making all natural 1’s and natural 20’s work the same for simplicity and clarity? Sure. That makes sense.

But, simultaneously, you’re adding a whole bunch of weird, nit-picky rules about which specific types of attacks and which specific types of character can get critical hits in combat?

That doesn’t make sense to me.

In any case, I was going to move on to explaining why the new auto-fail/auto-success rules for ability and skill checks isn’t as big a deal as you might think, but it quickly morphed into a wide-ranging discussion of why the 5^th Edition skill system is broken garbage built on top of questionable design principles.

So… buckle in.

It’s been awhile since I did a good ol’ fashioned D&D rant.

WHY BOUNDED ACCURACY?

Let’s start by talking about bounded accuracy. Endless ink has been spilt on this topic, but I think one of the clearest way to understand bounded accuracy — what it is, why it works the way it does, how it’s supposed to be used — is to look at the design lineage which created it.

To do that, we need to go back about twenty years to the development of the Epic Level Handbook for 3^rd Edition. The concept was to extend play past 20^th level, allowing players to continue leveling up their characters forever.

The big problem the designers faced was that different classes gained bonuses to core abilities — attacks, saving throws, etc. — at different rates, which meant that their values diverged over time. By 20^th level, the highest and lowest bonuses had already diverged so much that the difference exceeded the range of the d20 roll. This meant that any AC or DC you set would either be an automatic success for some PCs or impossible for others.

The designers of the Epic Level Handbook tried jumping through a whole bunch of hoops to solve or ameliorate this problem, but largely failed. As a result, the Epic Level Handbook was a pretty flawed experience at a fundamental level (and its failure may have actually played a major role in Wizards of the Coast abandoning the OGL and the doom of 4^th Edition, but that’s a tale for another time).

On that note, fast forward to 4^th Edition: The designers knew this was a problem. (Several of the designers had actually worked on the Epic Level Handbook.) They wanted to avoid this problem with the new edition.

Their solution was to level up everyone’s bonuses across the board: Classes would be strong at some things and weak at others, but the values wouldn’t diverge. This methodology was, furthermore, wedded to 4^th Edition’s design ethos of “level up the whole world with the PCs” and more or less fundamental to its My Precious Encounter school of encounter design.

Fast forward again, this time to 5^th Edition: The 4^th Edition of the game had burned down, fell over, and then sank into the swamp, and 5^th Edition’s mission was to win back the D&D players they had lost. The whole “level up the world” ethos was widely identified as one of the things people who hated 4^th Edition hated about 4^th Edition, so it had go.

Bounded accuracy was the solution. Importantly, bounded accuracy was about two things:

1. Controlling AC & DC so that the target numbers never become impossible for some of the PCs.
2. Controlling bonuses so that the results don’t become automatic successes for some of the PCs.

In other words, all of the results exist within that boundary. Hence, “bounded accuracy.”

If you go back to the original problem experienced in 3^rd Edition (and which metastasized in the Epic Level Handbook), you can see how this solves the problem. It also avoids the 4^th Edition problem where your numbers get bigger, but your results never actually improve because the numbers increase in lockstep: As long as the DCs remain consistently in bounds, the moderate increases to the PCs’ bonuses will see them succeed more often as they increase in level, resulting in high-level characters who feel (and are!) more effective than 1^st level characters.

BOUNDED ACCURACY & AUTO-RESULTS

This is also why my initial gut reaction to the new auto-fail/auto-success rules was wrong.

You don’t want a nat-1/nat-20 = auto-fail/auto-success rule in 3^rd Edition or 4^th Edition because the range of results shifts over levels and between characters: There are DC 35 tasks that you just can’t do unless you have a +15 bonus and that’s by design.

For years, in fact, I and many other people have preached the gospel here: Skill checks should not auto-fail on 1 or auto-succeeds on 20!

But bounded accuracy in 5^th Edition means that you should basically never be setting a DC that is impossible for one of the PCs to achieve. So having a natural 20 automatically succeed is irrelevant because it should already always be succeeding.

And if, in your opinion, a character should be succeeding on a roll of 1, then you shouldn’t be rolling those dice in the first place. You don’t make a Strength (Athletics) check to see if someone can walk across an empty street. Default to yes.

BUT BOUNDED ACCURACY IS BROKEN

What complicates this, however, is that bounded accuracy for ability checks/skill checks in 5E is broken.

The first problem is one of implementation: The instructions for setting check DCs are incorrect, which results in DMs setting DCs that break bounded accuracy.

The short version is that, for legacy reasons very similar to why I had my gut reaction to the playtest mechanics, the DC range in 5^th Edition is treated as if it were the same as low-level 3^rd Edition, including by the designers and the advice in the Dungeon Master’s Guide. But this isn’t the case. A skilled 3^rd-level character in 3^rd Edition likely has a +8 or +9 in the skill; the same character in 5^th Edition has +4 or +5.

Note: The elimination of the Take 10 mechanic in 5^th Edition for all practical purposes except passive Perception also has an effect here, but we won’t dive down that rabbit hole today.

This includes several key pieces of advice, which are given in various places throughout the 5^th Edition product line and reflected in the design of official scenarios and the like. (From here, this advice also percolates into designer diaries and third-party books, videos, tweets, blogs, etc.)

• DC 10 is the baseline “easy” check, relevant to unskilled characters.
• You should rarely or never call for PCs to roll for DCs under 10.
• You should step up your DCs by 5 points (going from DC 10 to DC 15 to DC 20).

The specific expression of this advice varies, but is fairly consistent. In the DMG, for example: “If the only DCs you ever use are 10, 15, and 20, your game will run just fine.”

But if you run the math, what you actually want is:

• DC 8 is the baseline “easy” check, relevant to unskilled characters.
• DC 12 should probably be your default difficulty.
• Thinking in steps of two is probably more useful: DC 8, DC 10, DC 12, DC 14, etc.

As I said, though, this is primarily a problem of praxis. In isolation, it could be trivially solved with better advice.

The more fundamental problem is mechanical: There are a handful of class abilities which trivially — but hilariously! — break bounded accuracy.

The rogue, of course, makes an easy example here. Expertise doubles proficiency bonuses, changing a range of +2 to +6 into a range of +4 to +12. Combined with ability score modifiers, this almost immediately turns most reasonable DCs within the system’s bounded accuracy into an automatic success for the rogue, and it gets worse from there.

Reliable Talent then comes in for mop-up, making the rogue’s minimum die roll 10. The rogue is now auto-succeeding on every proficient check, and in their chosen Expertise any DC that could challenge them is probably impossible for every other PC.

Of course, those are exactly the DCs these hilariously broken abilities pressure the DM to assign. Partly because they want to challenge the PCs. Partly because it just makes sense that these PCs should be able to achieve things the PCs without the hilariously broken abilities can’t do.

The end result, of course, is exactly the problem bounded accuracy was introduced to eliminate.

The new auto-fail/auto-success rules technically patch this up a bit:

• You’ve set a DC too high in order to challenge the hilarious broken character? At least the other PCs won’t auto-fail.
• You’ve set the DC “correctly” for bounded accuracy? The hilariously broken character can at least theoretically still fail.

But only in the crudest sense.

OPINION: THE TONE OF BOUNDED ACCURACY

So all of that is basically just math.

Now I’m going to digress into a purely personal opinion about why bounded accuracy makes the 5^th Edition skill system suck.

Let’s start by talking about why bounded accuracy works in combat: Hit points.

Although the typical Armor Class of a monster shifts upwards slightly as they increase in challenge rating, virtually every monster in the Monster Manual can be hit by a 1^st level character. This works, of course, because the amount of damage the monsters deal and the number of hit points they have do increase: The 1^st level character can’t readily defeat an adult red dragon because (a) the red dragon will smush them with a single attack and (b) the 1^st level character would have to hit them a bajillion time to whittle away their hit points.

The real advantage of this system is that it allows lower CR monsters to remain relevant: An encounter with twenty CR 1 dire wolves probably won’t threaten a 15^th-level party, but if you add them into an encounter with CR 13 storm giants, they won’t be completely irrelevant (since they can still hit the PCs and the PCs won’t necessarily auto-hit them).

But, of course, the skill system doesn’t have hit points or damage rolls. The “dire wolves” of the skill system never get easier to kill and you never become able to take on the “red dragons.”

As Rodney Thompson wrote in the article which introduced bounded accuracy to the world: “An iron-banded door is just as tough to break down at 20^th level as it was at 1^st.”

This creates a really weird dynamic where at 1^st level your characters struggle with dire wolves, casting dancing lights, and picking the lock on the back door of the tavern. And at 20^th level they’re soloing Smaug, summoning meteor swarms from the heavens, and… still having trouble picking that lock or kicking down that door.

And I don’t like it.

I recognize that there are other valid opinions here, but I would vastly prefer a skill system that unlocked abilities on par with all the other systems in the game (spells, combat, etc.). Having this weird, stagnant cul-de-sac creates some really bizarre effects in the fiction.

So what you’re left with here is a dichotomy. If you like the design principle on which bounded accuracy is built, you’re nevertheless left with the fact that 5^th Edition’s implementation of it in the skill system is hilariously broken.

And if, like me, you DON’T like the tonal implications of bounded accuracy in the skill system, then it’s just fundamentally undesirable AND broken.

THE SKILL LIST

Generally speaking, if you have a skill system in an RPG, then you want that skill system to be comprehensive. In other words, there should be skills covering the full gamut of tasks that PCs are likely to attempt. Or, to flip it around, you never want the GM to reach for a skill check and discover that the skill doesn’t exist.

Comprehensiveness should not be mistaken for minutia or complexity: You can achieve a comprehensive system by having forty different skills covering every sub-field of science, but you can also achieve it by just having a single Science skill. D&D 5^th Edition could hypothetically achieve it by having no skill system at all and just having characters be directly proficient in ability scores.

The 5^th Edition skill system, of course, broadly fails this basic criteria. I am constantly reaching for skill checks and then struggling to identify a skill which covers the task.

You can patch up some of these shortcomings by embracing the Skills With Different Abilities variant rule, in which skills can be paired to different ability scores depending on how they’re being used. For example, let’s say that you wanted to canvass a neighborhood for information. Non-variant 5^th Edition lacks any skill clearly covering that, but if you use the variant rule you can create a Charisma (Investigation) check and get what you need.

When you do this, however, you end up exacerbating another problem that I, personally, have with the system: Overlapping skills.

I vastly prefer a skill system in which I, as the GM, can call for a clear, definitive skill check. You may still end up with situations where players would like a different skill to apply (and you’ll need to make a ruling on that), but it’s a rare thing instead of affecting every single mechanical interaction in the game.

If you have a very large list of skills, the advantages of that expansiveness can, to some extent, justify the cost to precision if the skills end up with overlap. But despite having an incredibly short (and incomplete) list of skills, 5^th Edition still ends up with overlapped skills (e.g., Athletics and Acrobatics).

But okay, let’s lay my personal preferences completely aside: 5^th Edition has a short, concise skill list because it wants to keep the options streamlined. And it’s willing to accept the clunkiness of incompleteness to keep that relatively streamlined list.

Unfortunately, that’s when 5^th Edition slides up next to you and says, “Hey. Did I tell you about my OTHER skill system?”

Because, of course, it has one: Tool proficiencies. Massively overlapped (both with itself and with the skill list), not remotely streamlined, and more often confusing than not.

And also nonsensically crippled, because if you play according to the rules as written you can only make tool proficiency checks if you’re using the tool. So, for example, you can be a skilled carpenter, but that in no way translates to an ability to notice shabby construction, identify building materials, etc.

I’ll fully admit that, as far as I can tell, literally no one actually plays the game this way (including the designers), opting to allow this kind of knowledge-based tool proficiency check. But “a rule that nobody uses as written” is a pretty reliable indication of a rule that’s completely busted.

SKILL LOCK-IN

Finally, it’s fairly difficult to pick up additional skills in 5^th Edition. In fact, it borders on the impossible unless the DM is using other optional rules like feats. (I suspect the move in 1D&D to make feats non-optional and add more of them will help with this somewhat.)

This makes it quite difficult to adjust your character in response to the evolving circumstances of the campaign, something which skill systems are usually ideal for (since you can, in most such systems, make a multitude of adjustments during character advancement).

In the grand scheme of things, this is a fairly minor complaint. But if I’m going to write up a grand rant on all of my problems with the 5^th Edition skill system, I should at least try to be complete about it.

HOW WOULD YOU FIX IT?

For 1D&D? I wouldn’t. Backwards compatibility, in my opinion, is more important than tweaking the skill system.

Look, I told you this was a rant right at the beginning.

But if I had a time machine, could go back to 2014, and get a designer to listen to me:

• Make flexible ability score pairing the standard rule, not a variant.
• Eliminate the redundant skills.
• Add additional skills to provide a comprehensive skill list.
• Get rid of tool proficiencies.

And I’d make a strong case that bounded accuracy is the wrong call for the skill system and allow skill use to level up just like spell selection, combat efficacy, etc.

FURTHER READING

Untested 5E: Streamlined Skills
D&D: Calibrating Your Expectations