Back in 2007, I wrote D&D: Calibrating Your Expectations, an article which analyzed D&D 3rd Edition’s mechanics with a particular eye towards what characters of various levels were truly capable of.
(The short version: If you think Conan or Aragorn are 20th level D&D characters, then you’ll be disappointed and frustrated because 20th level D&D characters can routinely perform superheroic feats far beyond what Conan or Aragorn should be capable of. To avoid that dissonance, you need to either embrace awesomeness of high level play, cap your campaign at a lower level, or find an alternative solution for dialing in your personal sweet spot.)
A few years later, I wrote a similar article called Numenera: Calibrating Your Expectations, which did a similar breakdown for Monte Cook’s Numenera (and, by extension, the Cypher System).
These articles were fairly popular, and it’s perhaps unsurprising that I’m frequently asked to update my original article for D&D 5th Edition.
The reason that article has never been written is that, when you sit down and try to calibrate D&D 5th Edition, you’ll quickly realize that there’s nothing to calibrate: The D&D 5th Edition skill system, in particular, is a very shallow and hollow mechanic that is, at best, running on vibes.
The first thing I would do when attempting to calibrate a system looking at how a character’s action check results correlate to the real world. But the DCs in D&D 5th Edition aren’t pegged to anything in the real world: Walking across a balance beam could be DC 0, DC 5, DC 20, or DC 30. And the same could be true of running across the treetops while balancing on the leaves.
The next thing I’d do is compare the performance of skilled and unskilled characters. Particularly interesting values here include what characters consider routine (i.e., tasks they can automatically succeed at no matter what they roll on the dice) and also what tasks a skilled character can potentially achieve that an unskilled character cannot.
In D&D 5th Edition, however, skill proficiency bonuses are so tiny compared to both the range of the d20 roll and ability score bonuses that there’s no meaningful information to be gleaned. It’s quite likely, for example, that an unskilled character will actually have a higher bonus on a skill check than a character proficient in that skill, particularly at low levels.
This is, of course, intentional. For reasons I’ve previously discussed, D&D 5th Edition embraced the concept of “bounded accuracy.” The best way to think of bounded accuracy in this context is that everybody is basically equally skilled at pretty much everything, and stuff like natural ability and proficiency only indicate slight differences between characters. (It’s the difference between two people who went to the same university, got the same degree, and both graduated in the top 25% of the class, rather than the difference between a PhD in a subject and the average layperson.)
Bounded accuracy also pretty severely curtails the useful insights that can be gleaned from comparing characters of different levels. The differences between 1st and 10th and 20th level characters can be quite minute.
Oddly, though, D&D 5th Edition features a handful of class abilities which hilariously break bounded accuracy in ways which are almost absurd compared to the otherwise tiny differences in the system. I refer to these as Fuck Bounded Accuracy abilities, and these FBAs simply obliterate whatever marginal insight calibration might otherwise reveal.
So all you can really do is arbitrarily peg X to a DC, where X is some given task. But if you try to relate that back to a calibration question like, “What is a skilled vs. unskilled character capable of?” you’ll just end up frustrated because the system, again, doesn’t really distinguish one from the other.
CALIBRATING 5th EDITION
If you kind of flip this around, however, the formlessness of D&D 5th Edition offers you the opportunity to calibrate the system for yourself. In fact, it essentially demands this. Most DMs, however, only tackle this task blindly – groping in the dark and sort of arbitrarily throwing out DCs that “feel” right.
If you’d like to grasp this bull by the horns and instead make some conscious, deliberate choices about what you want the game to be, here’s what I would recommend.
First, define the maximum possible result you can imagine a 1st level PC ever achieving. Peg this as a DC 26 task. (Roughly speaking, this is a natural 20 (die roll) + 4 (ability score) + 2 proficiency bonus.)
Second, think about the minimum difficulty you would ever even remotely consider making a check for (and most of the time you probably wouldn’t). Peg that to DC 5.
Third, provide a spectrum of results between your DC 5 and DC 26 tasks. Keep in mind that at DC 17+ it’s technically possible to have a 1st level character who can NEVER accomplish those tasks. (If there were non-proficient and had a -4 ability score penalty.) DC 21, which is a task an average person could never achieve, is another useful benchmark to keep in mind.
Fourth, ask yourself what the most amazing thing you could ever imagine any PC doing at the highest levels of your campaign. Peg that at DC 36, then add spectrum values at DC 28 and DC 32 to bridge the gap.
Note that the tasks you assign to these DCs are entirely arbitrary, and will have a huge impact on the feel of your campaign.
For example, consider the ability to run across the top of the forest canopy like a character of Crouching Tiger, Hidden Dragon. This could just as easily be:
- Impossible
- DC 36 (most amazing thing you’ll ever see)
- DC 26 (a 1st level character can technically do it!)
- DC 18 (actually, it’s something everyone in Naruto can do)
Of course, if canopy running is DC 18 in your campaign, then you’ll need to figure out a DC 36 which is even more awesome than that. (Maybe being able to walk on clouds?)
Once you’ve calibrated your DCs, you need to accept that the FBAs will routinely allow PCs to punch way above the “intended” thresholds you’ve set. Consider this a feature, not a bug. Magic is awesome and the function of an FBA is to say, “Holy shit! Look at that guy!”
Finally, whatever you’ve set as your DC 36 touchstone is the cap. There’s nothing beyond that. If a player can string together some crazy string of bonuses and a natural 20 to roll a 52, that’s great. But the result will be no different than if they’d rolled a 36. They’ve reached maximum human potential, or however else you want to think of this. (This is necessary because the system’s issues – skill bonuses barely mattering; FBA abilities; etc. – make it impossible to calculate a coherent mathematical cap.)
Thanks for this. I came to pretty much the same conclusions a few years back, when I decided “fuck 5e” and went back to 3.0.
And don’t forget the mightily annoying Guidance cantrip, which makes the issue with skills even more ridiculous.
The other annoying aspect is that the FBA abilities create a dissonance such that characters are able to perform superheroic tasks but may fail miserably at mundane tasks, something we found completely counterintuitive. It’s like Superman can fly at supersonic speeds, but loses in a 100m running contest. WTF?
Is there a way to unbound the accuracy in 5e with a system similar to 3e, without actually running 3e?
This is something I ran into – without the system knowledge to explain – a lot while playing 5e, particularly because I tend to enjoy running low-level games and players tend to enjoy higher-level ones, so those aesthetic preferences would pull at each other. I’ve been thinking of running a short third tier game lately, and grappling with the scale is something I’ve been fumbling w/. I think sitting down and cooking up some alternative scales will be helpful.
Regarding Antonio’s complaint above, I sympathize, but ultimately don’t share the preference – I think it’s perfectly logical that characters with extraordinary abilities are still able to mess up! Video games like Disco Elysium simulate this pretty well; I’ve found Mork Borg fun for its swingyness for a similar reason. For me, it keeps a grounded “we’re just a bunch of guys” sense under all the flashy awesomeness.
Of course, this does make for a bad simulation of classic superheroics, but I’ve personally found that boring anyway :p
Nick – not really. 5e’s bounded accuracy assumptions are there in the way PC levels and monsters are designed, even if it’s loose. If you want a “treadmill” number goes up system, you need to use a 3.5/PF/PF2e style system that bakes those numbers in if you want any coherence.
@Nick On the basic level, unbounded accuracy 5e is pretty simple: increase the Proficiency Bonus numbers. The tricky thing is what to do with abilities that work around bounded accuracy (particularly Expertise since it’s the generic “this character/monster is skilled” ability) and if you’re using XP you’re going to have to do the monster math.
I’ve been really struggling for a while for how to set DCs in 5E that actually mean something in the context of my DMing style. The idea of calibrating your expectations for what each number kind of occurred to me but I had no idea where to draw the line between them. That latter section helps a lot!
The 5e system does at least differentiate more between different characters as level increases. An unskilled level 1 character may have +1 or +2 in a skill compared with +4 to +5 for a level 1 character with proficiency in a still using their main stat. Not a big difference when the distribution on a d20 is so wide and flat. We may as well all have a go at picking the lock. At 10th level the non-proficient character probably still has +1 or +2 and the proficient in a main-stat skill character now has +9 or +13 with expertise. A challenge appropriate to the proficient character is now DC20, very difficult for the non-proficient character, maybe only achievable 10% of the time and only worth attempting for minimal cost or major gain. At higher level the proficient character’s challenges become impossible for the non-proficient. This helps to distinguish characters more over time. Is it a problem that it’s not clear what the maxiumum performance is? Maybe not. Should it be possible to charm the Princess and persuade the King to arrange your wedding for tomorrow in one encounter? Maybe, stories go like this all the time and people build characters to be specialists. I think it’s good that the DM and players sort of get to work out at their table what is possible. I have played 2e games where more-or-less anything was possible from level 1 if it sounded cool and you rolled a 20. That’s probably worse than in 5e for realism but it’s lots of fun and engaged players with the situation – if you can think of a wacky solution you can try it. There are a few benchmarks in 2024 – Hide is DC15 even though it sounds like crouching down, throwing your cloak round yourself and being still and quiet for up to 6 seconds. Maybe that’s right, a typical commoner will have to be a bit lucky to go undetected, a proficient thief will mostly be successful from level 1 and with expertise will find this trivial by level 10. That feels about right in a heroic fantasy game to me, although is probably not representative of real-world human performance.
@Nick: You can change the dice used for non-combat or skill rolls, to 2d6 or 3d6. Because of how math works, this greatly reduces the chance element of a roll, since the results will cluster around the average. However, this then gets fucked by FBAs such as Expertise. But, it does get you closer, since (for example) the wizard will not try to do high DC physical tasks because they know they will probably fail.
@Nick – if you wanted to unbound the accuracy, I would take a page from Pathfinder 2e’s book and let players add their level to any skill they’re Trained in.
This is still going to be imperfect (PF2 gives players far more skill boost explicitly for this reason, plus has a wealth of other bonuses players will try to stack through magic items and feats, which are mandatory in PF2), but it will allow you to have a good gap between “what a level 1 peasant can do” with a +7 and “what you might expect a hero of the realms to succeed at” with a +28.
For a better solution, just dump 5e entirely. It’s not a good system on the whole, simply because it wants to be all things to all people. That lack of focus leaves the rules with a lot of gaps. Get yourself a good toolkit that runs the gamut for different system “feels” – a sci-fi game, a crunchy grid tactics game, a party game you can explain in five minutes for a wacky oneshot, a kid-friendly game you can run off the cuff, a heist system, etc.
It’s interesting how games that are all nominally “d20” can actually differ greatly once you scrutinize the mathematics.
Dungeon Crawl Classics RPG has an interesting (to me) skill system. While based off of the D&D 3.5e d20 system, it has no skill bonuses. Everything is just a check against one of the ability scores, but with the important provision that “untrained” characters (based on prior occupation/background) roll on a d10 and “trained” ones roll on a d20. Of course, the judge (GM) can also bump the roll up or down the “dice chain” for situational bonuses.
Sources of bonuses tend to be few, so DCs remain quite grounded. DC 5 is described as “child’s play”; DC 10 is a job for a professional (i.e., only trained characters will succeed with any frequency); DC 15 is a “feat of derring-do”; DC 20 is a truly heroic feat (e.g., Conan or Aragorn level skills).
In many ways DCC “runs on vibes” more than 5e (it is explicitly designed to provide a playing experience that emulates the sword and sorcery pulp literature that inspired D&D), but I have found that this skill system does a solid job of weighing training appropriately heavily and keeping DCs reasonably bounded to the expected values of the d20, plus typical bonuses.
I experimented with a variant in 5e which was basically, roll d20 for combat-time skills, roll 3d6 for knowledge/crafting skills, and 2d10 for anything inbetween (stealth, social skills, etc). The idea was, to use much the same rules and modifiers, but to make modifiers matter a lot more in situations where you’d expect an expert to be better than an amateur.
It seems useful, but I’m not sure how much difference it makes. There’s still the problem of too many skills with an odd +1 or +2 somewhere, which is theoretically supposed to feel like “intelligent character knows more stuff” but feels more random. I’d prefer characters to be able to specialise a bit more to give each character some different flavour, while still having a niche for “the intelligent one” and “the dextrous one”.
I find it odd that you ignore 5e’s own calibration of DCs from Very Easy 5 to Nearly Impossible 30. With 15 as Moderate, we see immediately that the calibration seems intended for someone “skilled in the art”, as patent lawyers put it. The typical level 1-3 PC with a +5 (a Fighter-1 with STR 16 & Athletics would count. A Rogue-1 with +7 Stealth from DEX 16 & Expertise is already more of an expert/master). Running across treetops IRL seems something only technically possible, so makes a good DC 30 Acrobatics if modelling the real world, though you could peg it at DC 25 (Very Hard) in a more Wuxia Physics campaign, in which case high level Rogues may be able to do it freely with Reliable Talent.
I’ve never set 5e skill DCs over 30, but if I did use 35 it would be for something literally impossible for a human to do IRL.
Samson bringing down the pillars of the Philistine temple seems like a reasonable DC 30 Athletics, something a STR 30 Epic Barbarian-20 can do automatically (but I’d likely model Samson as a mere STR 24 Barb-20 Raging and doing well on his +13 Athletics rolls with Advantage). I’d also use DC 30 Athletics for breaking open the Black Gate of Mordor. If I ever used a DC 35 it would be for something like shaking down the Tower of Barad-Dur. But for 5e I prefer to not allow actions beyond DC 30.
@Simon: Because the words “Easy” and “Nearly Impossible” aren’t actually calibrated to anything.
Easy for who? Nearly Impossible for who?
It can’t even be the same person, because to achieve a Nearly Impossible DC 30 task, you would need to have a +10 bonus. And if you have a +10 bonus, what is the distinction between Very Easy (DC 5) and Easy (DC 10) supposed to be?
This is very common in RPG rulebooks: It’s just a list of arbitrary numbers associated to arbitrary adjectives.
Justin, Simon:
On the one hand, I think that is roughly right for DC values.
On the other hand, I don’t think that’s at all the way to fix the problem, because the rules for resolution assume that five people who’ve never been to a city know as much about it as silly who lives there, which makes the DCs unsatisfying whatever value they get.
I think the bounded accuracy works reasonably well for combat, but the only plausible way to resolve the problems is to improvise resolutions for non-combat-time things.
Love this article and I think it will prove helpful for me, once I can get back to the gaming table. I did have one question about the below quote:
“DC 21, which is a task an average person could never achieve,”
Could you clarify what you mean by an “average person” here? Even the Commoner stat block with 10s across the board in Ability Scores has a +2 proficiency bonus. If there is something they are proficient in, then they can hit a DC 22 task with a natural 20. The same stat block doesn’t include skill proficiencies however, so the only thing it appears this theoretical person is good at is hitting people with clubs.
Were you basing the above statement on the “average person” not having skill proficiencies as well as not having any ability score bonuses?
@Nathan: Basically, being skilled/proficient at something would mean you’re not an average person when it comes to that task.
Now, any specific person you pluck off the street might succeed on that task because they’re above average for any number of reasons (they have help, they have a higher ability score modifier, etc.). But, statistically, they’d be just as likely to be below average and actually incapable of, say, a DC 18 task.
It’s a useful benchmark, not some hard-and-fast rule about a Platonic average person.