What is a "fair" DC?

General questions, debates, and rants about RPGs

Moderator: Moderators

Post Reply
Mord
Knight-Baron
Posts: 565
Joined: Thu Apr 24, 2014 12:25 am

What is a "fair" DC?

Post by Mord »

For a while now I've been working on the math behind a 2d10 RNG used to power a 20-level RPG system. I have a few cool spreadsheets now which contain all the number crunching I've done so far as well as some notes: Googledoc link.

But while the spreadsheets are a good tool for exploring the implications of various choices, at this phase I'm interested in interrogating the ideas that underly all the math; I'd like some feedback to see if these underlying concepts pass the sniff test for anyone who isn't me:

1. I think the default DC for a level-appropriate check should give the character of that level with the lowest bonus the lowest success chance not less than 50%.

2. I think two characters are only the same "level" in a meaningful way if there is no check that one cannot pass that the other simultaneously cannot fail.

3. I feel like there's not a lot of point in setting DCs such that the worst character has <10% success rate or the best has a >85% rate.

4. I also feel like there's not a lot of point in allowing for bonus spreads so great that anyone in the party has a 100% chance of success on rolls that are at the default DC for their level.

Based on these premises, I'm pretty happy with the way the progression matchups stack up, where a character of level X on progression Y has a Z chance of succeeding versus a target at level W on progression V.

However, I foresee that the tight constraints on bonus spread may cause problems when it comes time to build the actual character classes and spells, while the tight DC range will make it difficult to design adventures with a meaningfully diverse set of challenges.

So I guess my broad question is: am I being too restrained in my assumptions of what is a "fair DC?"
Schleiermacher
Knight-Baron
Posts: 666
Joined: Wed Sep 05, 2012 9:39 am

Post by Schleiermacher »

In my opinion, this framing of the problem is backward. The only way to have a system that describes a consistent, verisimilar setting rather than a 4e-style bonus treadmill is to design the challenges first and determine what's level appropriate at any given level from there, rather than vice versa.

In terms of class design I feeel like the most important premise is premise #2, the others are more about calibrating the numbers, which IMO you should do by regression analysis after you've written up the challenge space, see above. As such I don't care a lot about most of your premises.

Premise #2 is a little interesting though. If you want an action resolution system where character level is one of the most important determinants of skill, such that high-level Cavaliers routinely out-sneak low-level Assassins, premise #2 is a good way to define what "level" and "level appropriate" means. But:

1. then you need to be aware of what you're writing and the setting implications thereof.

And 2. that's not the only valid or functional way to define what level-appropriateness means.

It's fine for characters who work in different idioms to have different skill sets, even to the point that one auto-succeeds where another auto-fails, as long as that doesn't lock either of them out (or allow them to opt out) of being able to participate meaningfully in core gameplay, whatever that means for any given game.

It's totally fine in the abstract for an adventure to have a Stealth challenge that the Assassin in the party can't fail and the Cavalier can't pass, as long as passing that challenge isn't the only way to succeed in the adventure. (In such a case the challenge wouldn't be "level appropriate" after all since the Cavalier can't pass it, and more importantly, it's restrictive, overly linear and generally bad adventure design.)
Last edited by Schleiermacher on Mon Apr 10, 2017 9:33 pm, edited 1 time in total.
User avatar
tussock
Prince
Posts: 2937
Joined: Sat Nov 07, 2009 4:28 am
Location: Online
Contact:

Post by tussock »

So, I can't help but notice, aside from the assumption of a level treadmill, you make no accounting for retries, group attempts, multiple characters all attempting separately, or any of that.

#1: 45% failure is happening constantly, that's certain failure for anything that can ever matter.

#2: You seem to be OK with automatic success, and with effectively automatic failure, but not with both at the same time? That's, um, picky. Did you maybe just read that somewhere and not think it through? Like, maybe there are mostly characters other than the perfectly fit to task and not at all fit.

#3: If you're only rolling once, a 10% success rate never works, likewise, 85% success rates fail all the time if you roll more than once, and still a lot at just once. Assuming you roll these things more than once in a session, I hope the game can handle the quantity of failures you're throwing at perfect experts and stuff.

#4: Really? You're not letting perfect experts succeed all the time at basic "level appropriate" tasks? That's not 100%? That's not, like, 100% after massive penalties?


So, like, OK, but maybe consider ...

What does failure even mean? Setbacks, damage, more things needing done? If failure is "try again" then none of this even matters, just roll a d6 until you get a 6. If failure does hurt, and will stack with other failures, then fail rates need to come way down.

If you're allowing multiple characters to try to succeed, what happens? Because you could be collapsing the failure rates to near-zero if they all have to fail, or collapsing the success rates no near-zero if they all have to succeed, or doing unexpected things if half of them have to succeed (seriously, it'll flick from near-0% to near-100% very quickly).


And if you need to chain these things together to get anywhere, hello, that's also way too low of a success rate. So, what are you even doing with these numbers, man?
PC, SJW, anti-fascist, not being a dick, or working on it, he/him.
User avatar
Chamomile
Prince
Posts: 4632
Joined: Tue May 03, 2011 10:45 am

Post by Chamomile »

Building a game math first is bonkers. The point of the math is to effectively simulate whatever kinds of stories you're trying to tell. You should be figuring out a diverse set of adventures first, then determine what math is necessary to modeling those adventures.
Mord
Knight-Baron
Posts: 565
Joined: Thu Apr 24, 2014 12:25 am

Post by Mord »

Schleiermacher wrote:In my opinion, this framing of the problem is backward. The only way to have a system that describes a consistent, verisimilar setting rather than a 4e-style bonus treadmill is to design the challenges first and determine what's level appropriate at any given level from there, rather than vice versa.
How do I design a challenge that is appropriate for characters of a given level when I don't know the DCs that characters at that level are capable of meeting, and how likely it is for them to do so?
Premise #2 is a little interesting though. If you want an action resolution system where character level is one of the most important determinants of skill, such that high-level Cavaliers routinely out-sneak low-level Assassins, premise #2 is a good way to define what "level" and "level appropriate" means.
That is entirely what I'm about, yes, so this is a good sign.
But:

1. then you need to be aware of what you're writing and the setting implications thereof.


And 2. that's not the only valid or functional way to define what level-appropriateness means.
Sure. What are some of those implications, as you see them?
It's fine for characters who work in different idioms to have different skill sets, even to the point that one auto-succeeds where another auto-fails, as long as that doesn't lock either of them out (or allow them to opt out) of being able to participate meaningfully in core gameplay, whatever that means for any given game.
Sure - not everyone can or should have max ranks in Hot Dog Eating, so when the Hot Dog Eating contest happens, the guy who specced it gets to feel like he didn't waste his time.

That said, I should have clarified that I'm considering these progressions for things akin to "save DC versus save bonuses" and "BAB versus AC" rather than skill checks, because I'm trying to work out the core gameplay mechanic first.
It's totally fine in the abstract for an adventure to have a Stealth challenge that the Assassin in the party can't fail and the Cavalier can't pass, as long as passing that challenge isn't the only way to succeed in the adventure. (In such a case the challenge wouldn't be "level appropriate" after all since the Cavalier can't pass it, and more importantly, it's restrictive, overly linear and generally bad adventure design.)
One of my assumptions, that auto-results are less fun than having to roll for it, is directly opposed to your scenario here, so I am really interested in your reasoning. Do you find it fun when no one at the table needs to roll for something that is supposed to be a major scripted challenge due to good character building or preparation or whatnot? For my part, I'm imagining here the opening scene of Raiders of the Lost Ark, except instead of cautiously weighing out his bag of sand and making the quickest switch he can, Indy just yoinks the idol and strolls off with it scot-free.

In my opinion, it's OK if Princess Leia and Han Solo have literally zero chance of hitting the Death Star exhaust port from an X-Wing cockpit, since there's still some tension in the scene as long as Luke isn't guaranteed to hit it either and the others can participate in the scene in some different way.

tussock wrote:So, I can't help but notice, aside from the assumption of a level treadmill, you make no accounting for retries, group attempts, multiple characters all attempting separately, or any of that.
Those are good things to consider regarding skill checks and such, but these progressions are first and foremost meant to handle the nuts and bolts "BAB vs AC" kind of mechanics.
#1: 45% failure is happening constantly, that's certain failure for anything that can ever matter.

[...]

#3: If you're only rolling once, a 10% success rate never works, likewise, 85% success rates fail all the time if you roll more than once, and still a lot at just once. Assuming you roll these things more than once in a session, I hope the game can handle the quantity of failures you're throwing at perfect experts and stuff.
OK, we get to the meat of it. I picked my default 2d10 DC of 11 based on the 3.X average DC of 10; both of these have a 55% chance of success for a character with +0 bonus. Do you feel that 3.X also should have lower DCs in general? What, in your opinion, is the ideal chance of success on an "Average" check (or a basic attack vs a CR1 monster) for a rando lv1 adventurer?
#2: You seem to be OK with automatic success, and with effectively automatic failure, but not with both at the same time? That's, um, picky. Did you maybe just read that somewhere and not think it through? Like, maybe there are mostly characters other than the perfectly fit to task and not at all fit.
To the contrary: I don't like setting DCs such that the worst character has les than 10% chance to succeed, nor do I like setting DCs such that the best character has greater than 85% chance to succeed. Nor do I like the idea of someone having such a high bonus compared to the rest of the party that they auto-pass checks at their level's default DC.
#4: Really? You're not letting perfect experts succeed all the time at basic "level appropriate" tasks? That's not 100%? That's not, like, 100% after massive penalties?
I would define a "basic" task as one that has a relative DC in the 7-10 range. On such tasks, experts would have a 90%-99% success rate, which I think checks out.

F'rinstance: your Assassin with a +4 to Reflex Save has a 99% chance to hit a DC 7 avoid-the-crappy-poison-dart-trap check, while your Wizard type with +0 to same has an 85% chance to do the same. If anything I feel as though the wizard is getting too good a deal here, and that therefore you as the MC should only call for a roll if the DC is that low when you want everyone to pass - perhaps this DC 7 check comes in when the party triggers a low-power firebolt trap in the first rooms of a low-level dungeon.

For a DC 11 dodge-swinging-scythe-blades check, Joe Assassin is now batting at 85%, while Jimmy Wizard is down to 55%.

The examples above seem reasonable to me, but again, I'm looking for your feedback - is this too low?
So, like, OK, but maybe consider ...

What does failure even mean? Setbacks, damage, more things needing done? If failure is "try again" then none of this even matters, just roll a d6 until you get a 6. If failure does hurt, and will stack with other failures, then fail rates need to come way down.
Good question. My current thinking on the subject is that a failure that misses the DC by 1-9 points has no special consequences beyond losing the time it took to make the check (may or may not have a chance to retry depending on circumstances), whereas a failure that misses the DC by 10+ does bring with it some kind of consequences - something like "you broke the lock by doing such a shitty job of picking it" or "you committed a terrible faux pas and now the Duke won't talk to you".
If you're allowing multiple characters to try to succeed, what happens? Because you could be collapsing the failure rates to near-zero if they all have to fail, or collapsing the success rates no near-zero if they all have to succeed, or doing unexpected things if half of them have to succeed (seriously, it'll flick from near-0% to near-100% very quickly).

And if you need to chain these things together to get anywhere, hello, that's also way too low of a success rate.
That is a serious problem, but it's not unique to any RNG or set of guideposts on an RNG. It's easy enough to deal with all this when checks are made 1 character at a time. Going back to talking about skill checks, your Assassin might silently scale castle walls for breakfast, but when he's dragging along Cavalier, Wizard, and Swordguy, he's basically hopeless.

The options, as I see them, are:
  • keep the spotlight on character specialties as briefly as possible (not satisfying to the person who specced Hot Dog Eating when the hot dog contest is one roll and only happens once),
  • split the party in order to put the spotlight on everyone's specialty in turn (doesn't work because everyone else goes to play Smash until their turns),
  • or give everyone a chance to contribute in all circumstances... or to at least be relatively assured of not fucking it up for everyone.
I feel like Door #3 there is the least of those evils, but maybe there's a Door #4 I'm not seeing. Perhaps in cases like Assassin trying to drag the whole party up the wall with him, the best solution might be to cover for his partymates' lack of skill by taking a different approach. I'm imagining a zany scheme involving fake beards and a work order for latrine cleaning; something that would still leverage Assassin's stealth abilities by letting his skills give a benefit to his teammates while still letting them participate.
So, what are you even doing with these numbers, man?
A good question. Ideally, making a game about gaining power by beating up monsters and using that power to beat up bigger monsters until you find yourself punching Zeus in the face because you don't like how he runs the cosmos. Realistically, probably nothing.
User avatar
Chamomile
Prince
Posts: 4632
Joined: Tue May 03, 2011 10:45 am

Post by Chamomile »

Mord wrote:
Schleiermacher wrote:In my opinion, this framing of the problem is backward. The only way to have a system that describes a consistent, verisimilar setting rather than a 4e-style bonus treadmill is to design the challenges first and determine what's level appropriate at any given level from there, rather than vice versa.
How do I design a challenge that is appropriate for characters of a given level when I don't know the DCs that characters at that level are capable of meeting, and how likely it is for them to do so?
With words. Once you have a concept of what it means to be level 6, you can figure out what kind of numbers people need at level 6.
Username17
Serious Badass
Posts: 29894
Joined: Fri Mar 07, 2008 7:54 pm

Post by Username17 »

The amount of failure that is acceptable is determined by how big a setback any particular failure is. If you lose the evening on a failed roll, having it happen even one time in ten is nutzo. If you have to wait five minutes to roll again, then you're looking at it being something that should happen less than once in 10 times to fail three times in a row, and a 45% fail rate is OK. The question of how frequently an attack can miss is unanswerable without filling in more information. In a game like Diablo or Call of Duty, it's acceptable for a majority, even a large majority of attacks to miss. You're gonna get another attack the next time you hit the space bar. In a monthly play by mail game, missing is simply not acceptable.

But the more basic question is one of what people should succeed at doing, and once that has been done you can decide how frequently people should announce their intentions between here and there.

-Username17
Pedantic
Journeyman
Posts: 125
Joined: Thu Feb 09, 2012 12:42 pm

Post by Pedantic »

If you're just looking at combat math, that's very different than skill rolls.

For skill checks, there's other options than just keeping everyone so tightly in lockstep. Consider mechanics that let players share or swap bonuses (Fantasy Craft has a class that does this, and it feels pretty good because the specialist player feels like they've made a big contribution) or group check mechanics.
User avatar
OgreBattle
King
Posts: 6820
Joined: Sat Sep 03, 2011 9:33 am

Post by OgreBattle »

For D&D style dungeon crawling I like rolling two attacks with a 66% chance for any given hit to land. I have a high chance of landing at least one hit, while landing two hits feels special enough to satisfy my desire for degrees of success.
User avatar
Occluded Sun
Duke
Posts: 1044
Joined: Fri May 02, 2014 6:15 pm

Post by Occluded Sun »

The question you have yet to bring up, and which is absolutely necessary, is this: How much difference should level make?

Should a level 10 character utterly outclass a level 1, or be only slightly superior? Should a level 10 character be able to do things a level 1 character has no chance at, or are roughly the same things possible or impossible?

If you don't know the answers to those questions, there is no basis for answering yours.
"Most men are of no more use in their lives but as machines for turning food into excrement." - Leonardo di ser Piero da Vinci
GâtFromKI
Knight-Baron
Posts: 513
Joined: Fri Sep 02, 2011 10:14 am

Re: What is a "fair" DC?

Post by GâtFromKI »

Mord wrote:3. I feel like there's not a lot of point in setting DCs such that the worst character has <10% success rate or the best has a >85% rate.
With this requirement you can use a d6 as your rng ; 1 is auto-fail, 6 is auto-success.

Level-appropriate DC is 3. It's a success more often than not, but it seems to fail constantly (33% is a huge failure ratio).

A strong relevant attribute gives +1; an appropriate skill gives +1. Therefore with the right attribute and skill, you success at very hard checks on 2+. A weak attribute gives -1 (skills can't go below "untrained", they can't give a penalty).


You can use another RNG for more granularity, anyway "2/3 chance of success" is a good start for "level appropriate challenging DC"; and "+1/6 success (~17%)" is a good start for "a bonus large enough people are caring about". If you want a minimum success/failure rate of 15%, I don't see any reason to not use the d6 as your RNG.
Last edited by GâtFromKI on Fri Apr 14, 2017 12:29 pm, edited 1 time in total.
Mord
Knight-Baron
Posts: 565
Joined: Thu Apr 24, 2014 12:25 am

Post by Mord »

Based on most of the feedback I've received - and I appreciate the responses! - it's clear that I'm approaching this from a direction that is at best orthogonal to the way the rest of y'all are, or at worst ass-backwards.
Chamomile wrote:With words. Once you have a concept of what it means to be level 6, you can figure out what kind of numbers people need at level 6.
One recurring theme in criticisms is that I should start with designing the challenges and then work out the math. Fair enough. On the scale of being, I see your standard high fantasy and Greek heroic adventures being carried out from levels 1-5, growing into Spelljammer or Planescape territory in the 6-10 band. Levels 11-15 are mythic demigod tier, with levels 16-20 being occupied by literal Gods and the hidden bosses in Final Fantasy games.
FrankTrollman wrote:The amount of failure that is acceptable is determined by how big a setback any particular failure is. If you lose the evening on a failed roll, having it happen even one time in ten is nutzo. If you have to wait five minutes to roll again, then you're looking at it being something that should happen less than once in 10 times to fail three times in a row, and a 45% fail rate is OK. The question of how frequently an attack can miss is unanswerable without filling in more information. In a game like Diablo or Call of Duty, it's acceptable for a majority, even a large majority of attacks to miss. You're gonna get another attack the next time you hit the space bar. In a monthly play by mail game, missing is simply not acceptable.

But the more basic question is one of what people should succeed at doing, and once that has been done you can decide how frequently people should announce their intentions between here and there.

-Username17
To clarify, I'm assuming your bog standard face-to-face setup where you've got 1 MC and 3-6 PC characters. Ideally I'll find ways to streamline the core game systems (foremost of which is combat) such that each PC really can expect to get 1 turn in every 5 minutes. That said, one of my many thus-far-unspoken ideas is to eliminate "missing" pretty much entirely, in favor of having something that would otherwise be a "no effect, you accomplished nothing, go back to Snapchat" miss instead be a "glancing blow, you don't do full damage but depending on the move you did you might still inflict a status effect, do a little damage, recharge your power pool, or otherwise get something out of it in some way so you don't feel like this was a total waste of time."

It's now apparent to me that attempting to divorce my math ideas from the rest of my ideas for the purpose of soliciting feedback was a mistake. My bad.
Pedantic wrote:If you're just looking at combat math, that's very different than skill rolls.

For skill checks, there's other options than just keeping everyone so tightly in lockstep. Consider mechanics that let players share or swap bonuses (Fantasy Craft has a class that does this, and it feels pretty good because the specialist player feels like they've made a big contribution) or group check mechanics.
Indeed. Your comment inspired me to take a closer look at the iterative probabilities of party skill checks, and it strikes me just how drastic the difference is between a "one person must pass this check" (e.g. reading the mysterious inscriptions on the wall) and an "all of you must pass this check" (e.g. doing anything undetected as a party). For a party of 4 with no bonuses, you can expect to pass a DC16 check of the former type on a d20 68% of the time, while getting 66% odds of success on a check of the latter type requires you to drop the DC to 3... ouch. Perversely, having a specialist in that party of 4 helps a lot on the former type of check, where they really weren't needed, and virtually not at all on the latter type, where they really would be.

Allowing specialists to share their bonuses with the party or substitute their bonus (or otherwise compensate for their doofus partymates' failures) is a really good idea, and something that should come online for every class very early in the game. Or, alternatively, this could also be considered a symptom of the need for more fleshed-out minigames to handle things that normally tempt an MC to call for an "everyone must pass" check. Whatever the case, 4e skill challenges are a cautionary example.
OgreBattle wrote:For D&D style dungeon crawling I like rolling two attacks with a 66% chance for any given hit to land. I have a high chance of landing at least one hit, while landing two hits feels special enough to satisfy my desire for degrees of success.
Excellent; noted, thank you. :mrgreen:
Occluded Sun wrote:The question you have yet to bring up, and which is absolutely necessary, is this: How much difference should level make?

Should a level 10 character utterly outclass a level 1, or be only slightly superior? Should a level 10 character be able to do things a level 1 character has no chance at, or are roughly the same things possible or impossible?

If you don't know the answers to those questions, there is no basis for answering yours.
That's a complicated question that requires additional inputs in the form of "how good is each character at the thing they are opposing on," since, for instance, two characters on the Bad progression for the contended roll will need a larger level gap for one to utterly dominate the other than one Bad character and one Perfect. But absent that specificity, being as general as possible, it should take between 8-16 levels of difference (call it 12?) for the higher-leveled character to be able to squash the inferior like a bug while running no risk of being meaningfully harmed in the process.

As for "what should a level 10 character be able to do in relation to a level 1 character's ability," I feel pretty good about your hypothetical Level 10 character being able to do 100% of the time something that your Level 1 character is able to do 64% of the time.

To put some concrete numbers on that: imagine a Good level 1 character has a +1 bonus to do Thing X, which means he can succeed on a DC 20 check 3% of the time. A Good level 10 character has a +8 bonus to do Thing X, which means he can succeed on a DC 20 check 45% of the time. The level 10 character can also attempt checks with DCs ranging from 22-28 with a nonzero chance of success, all of which are flat-out impossible for the Level 1 character. Indeed, the L1 guy's best trick, which is passing a DC21 check 1% of the time, is something the L10 guy can do 36% of the time. Maybe that isn't good enough? By L20, though, that guy has a +16 bonus and is able to hit DC21 94% of the time; it just ain't even a thing.

So, to put things in some kind of perspective - by the end of his life, Achilles was probably no higher than L8. Paris, who was basically some minor Trojan princeling (L3?) could still slay him, though only with specific knowledge of his weaknesses and not in a stand-up fight. Hercules, on the other hand, having ascended to Godhood (possibly even Olympian Godhood depending on who you believe) would be somewhere in the L15 range, meaning that Paris would no chance at all against him.
GâtFromKI wrote:
Mord wrote:3. I feel like there's not a lot of point in setting DCs such that the worst character has <10% success rate or the best has a >85% rate.
With this requirement you can use a d6 as your rng ; 1 is auto-fail, 6 is auto-success.

Level-appropriate DC is 3. It's a success more often than not, but it seems to fail constantly (33% is a huge failure ratio).

A strong relevant attribute gives +1; an appropriate skill gives +1. Therefore with the right attribute and skill, you success at very hard checks on 2+. A weak attribute gives -1 (skills can't go below "untrained", they can't give a penalty).

You can use another RNG for more granularity, anyway "2/3 chance of success" is a good start for "level appropriate challenging DC"; and "+1/6 success (~17%)" is a good start for "a bonus large enough people are caring about". If you want a minimum success/failure rate of 15%, I don't see any reason to not use the d6 as your RNG.
That's a good point, and if the game were meant to be played at a single power bracket (e.g. Shadowrun, World of Darkness) this would be a good move, but there are 2 reasons I want to stick with the 2d10: 1) the RNG needs to be able to handle gradations smaller than 17% in order to model characters progressing across 20 levels in a smooth manner (and sometimes meeting challenges outside their own power level), and 2) curved RNGs have an increased central tendency as compared to flat, which I find useful.
Post Reply