A Short Rest: Gnomish Workshop: Challenge Ratings… We need a bigger analyzer!

A Short Rest: Gnomish Workshop: Challenge Ratings… We need a bigger analyzer!

This article was first broadcast in Episode Fifty-Two on 12th December 2018.

RyuOh no he’s down! Ostron!
LennonDid the KDM do this?
RyuHey! I haven’t worn the hat for…a while now and besides he doesn’t look hurt. He’s sort of more in a trance I think? What’s this thing he’s lying next to?
LennonWell it’s…wow. It’s taking up a lot more room than I thought at first glance. Oh there’s a label! “Realized Oikumenatal Solution To Removing Obfuscation”
ROSTROYou may refer to me as ROSTRO
Ryu & LennonGAAAAH!
RyuWhat did you do to Ostron?
ROSTRONothing. Technically, he did something to me.
Lennon…and what was that?
ROSTROMy activation required additional cognitive energy, therefore he astrally projected himself into my mechanisms.
RyuHow is that even possible?
ROSTROMy internal workings are in practice a small artificial plane formed from mathematical probability.
LennonThat sounds…terrifying
ROSTROOstron determined it was necessary for the research project he was pursuing.
RyuWhat was that?
ROSTROPlease observe the output crystals


From the beginning of 5th edition, the Challenge Rating of monsters presented by WotC has been a subject of criticism and debate. But before we start getting into the utility of Challenge Rating (or CR as it’s usually written), it’s worth examining what CR was trying to accomplish. For that, we again have to look back at the history of D&D.   

Until the inception of 5th edition, creatures were assigned levels similar to the ones characters had. Many humanoids even acquired specific classes and the abilities commensurate with said classes. A hint of this remains in 5th edition when you come upon a creature capable of using magic; each one will be labeled as an “nth” level spellcaster with access to a particular spell list.

In previous editions, the utility of this nomenclature…okay no, sorry, Lennon, did you program how this thing talks? I’m summarizing! While being able to figure out why creatures had certain abilities was sometimes nice, there were two mechanical reasons the creatures had levels: scaling AC, and magic.

Yeah, definitely summarizing this gibberish.

Scaling AC was the big one. In most editions and especially in 4th, players’ ACs were somehow increased based on the characters’ levels. In some cases this also applied to attack bonuses. So if a player and a creature of wildly different levels ran into one another, the lower level creature would be literally unable to hit the higher level one. So the levels helped DMs determine which creatures would pose actual threats to players, and which ones could be safely thrown in the players’ way without any worry of the players killing it if the DM didn’t want.

The other implementation, spellcasting, has largely been eliminated from 5th edition, though a snippet remains in the mechanics of the Turn and Destroy Undead Cleric class features. In previous editions, several spells and spell effects considered the level of targeted creatures, as opposed to checking a defense or forcing a saving throw.

With the introduction of the bounded accuracy system to 5th edition, the primary mechanical reason for assigning levels to creatures was eliminated: Armor Class values no longer adjusted based on level. Any creature of any level can, in theory, successfully attack any other creature in the game. Consider this example.

RyuWait…no pick another one.
RyuAnother creature. No, another. Oh come on, you can analyze things, figure out where I’m going here.
ROSTROExtrapolating…Very Well

Let’s look at Tiamat, because it’s me and she’s a dragon. In 4th edition, Tiamat was statted as a Level 35 creature with an AC of 51. Characters couldn’t even think about successfully attacking her until at least the mid-20 level range. But now take 5th edition: the dragon god is a CR 30 monster, with an AC of 25. A level 1 character with a +4 Strength modifier, and a weapon they’re proficient with, technically has a chance of landing a hit against the dragon queen if they roll a 19 or 20. I mean, they’ll be vaporized as soon as Tiamat takes a turn, but mathematically, if you can get an army of 1000 peasants to all take a swing at the dragon god, 100 of them get to say they got a hit in before they’re all eaten.

Since there is no more “level gating” before characters can engage other creatures and spells aren’t using levels for their effects anymore, creatures didn’t need to have levels. But the levels helped people figure out if they could fight the creatures, right?

The problem was, and some say still is, if you aren’t tying the levels to defense or damage, level doesn’t equate to danger. Even for player classes, a level 9 bard and a level 9 barbarian represent wildly different threat levels as combatants. But now that players can technically hit anything, WotC needed something to help DMs produce balanced encounters other than saying “average the damage your creatures can do and their AC, then compare it against the player’s HP totals and their bonuses to hit.”

The Challenge Rating system is wizards’ attempt to codify the mathematical comparisons that theoretically determine the likelihood of players’ success in confronting creatures of differing capabilities. Pages 273 through 281 of the Dungeon Master’s Guide have been referenced before on this production, and they provide guidelines around the numerous variables affecting a creature’s challenge rating. However, the most significant referential information is located on page 274: the “Monster Statistics by Challenge Rating” table.

This table exemplifies the idea that challenge ratings are primarily based upon a combination of 5 factors: Armor class, total hit points, Attack bonus, damage output per round, and the required DC for any save effects the creature imposes. Proficiency bonus is also listed, however that is rarely considered apart from the attack bonus on creatures, so it was discounted.

So if the table is a good guide, a CR 1 creature should have an AC of 13, a hit point total between 51 and 85, a +3 attack bonus or a save DC of 13, and do 9 to 14 damage a turn. Now there is some wiggle room described in the rules. A higher AC can be offset by a lower hit point total still resulting in a similar CR. Or a low AC, low HP creature will have it’s CR increased because it hits like a truck. But in general we know the ranges we should be looking at.

The problem comes if you try to apply those rules to most of the published creatures in official sources. ROSTRO says Ostron pulled a random sample of five creatures and attempted to justify their CR rating by using the method from the DMG and none of them even came close.

Lennon: Hey, wouldn’t you be capable of doing, like, hundreds of comparisons like that?
ROSTROMy faculties were engaged in more worthwhile pursuits.
ROSTROI am not at liberty to discuss that at this time
RyuOh, I’m not worried at all now

In addition to failures in applying the prescribed mathematical comparisons, anecdotal evidence from players and DMs suggested that creatures very often performed below or above their expected outputs based on their challenge ratings.

Theories abound as to the reasons. My own analysis suggests a primary factor in the inaccuracy is the disregard of the system for the impacts of action economy. The Roper is a creature oft cited by Ostron when Action Economy is discussed, but with good reason as it presents the clearest example. The Roper acquired a CR of 5 and the numerical comparisons to the Dungeon Master’s Guide table more or less corroborate the assignation. However, the CR table ignores the fact that the Roper has the ability to restrain up to four creatures at a range beyond any melee and many ranged attackers.

Using the encounter danger math suggested by the DMG, one roper vs 5 level 5 characters should be an encounter with easy-to-medium level danger. But when the roper restrains three of the characters 60 feet from it and drags another one into close combat, suddenly it’s more like 2 level 5 characters. So the encounter’s gone from easy-to-medium up to hard-to-deadly after one round of the monster attacking.

As ROSTRO said, the Roper is just the most obvious example, but other examples abound, such as the Gorgon that can restrain and petrify creatures or the Bodak that can, as a CR 6 creature, insta-zero any creature that fails a save badly enough and it doesn’t even spend an action for that; it just happens. None of these creatures are Legendary. The CR system doesn’t seem to have any way to account for these abilities and how they affect the difficulty of combat encounters.

These shortcomings have led many people to conclude the challenge rating system is at best a loose guideline and at worst serves no purpose whatsoever. In casual settings this has resulted in many DMs opting for the so-called “Milestone leveling” system where subjective accomplishments trigger level advancement rather than accrual of numerical experience based on vanquishing foes. Some, particularly my creator, have even suggested the recent change in policy in leveling for the Adventurer’s League is a direct result and tacit acknowledgement by Wizards of the Coast that the CR system is not a reliable measure of encounter difficulty.

The reality is reliably measuring the difficulty of an encounter apart from experimental simulation is problematic; too many variables exist in practice. Despite Ostron’s comparisons, many monsters do numerically equate to their proper CR once their stats are calculated. Most of the problems derive from creatures’ other abilities. It is therefore recommended that DMs use creature CRs as a starting point, but the actual abilities, armor class, and damage output of creatures should be reviewed and compared to player’s statistics and capabilities before committing. For example, have the creatures simply attack without employing any special abilities, then introduce said abilities if the difficulty level seems low. I hope this informative analysis has been sufficient to your needs.

Sounds of machine shutting down
RyuUmmmm…did it just shut off?
LennonI…I think so.
RyuDoes that mean…um….
LennonOh I really hope not
OstronUm, hi?
RyuOh! Ostron!
OstronYeah, sorry I was working on a thing. Oh good did the Realized Oikumenatal Solution To Removing Obfuscation give some useful output?
LennonI guess you could say that, sure.
OstronExcellent! Looks like the magic power cell ran out just after…Oh look at the time, why are you guys just standing here? Scrying pool!
Ostron walks off
RyuDoes he know he was… you know…?
LennonDoesn’t seem like it
RyuDo we tell him?
LennonLet’s not for right now. I’ll talk to him later. For now let’s just meet him at the Scrying pool.