DC 10. You roll a natural 1, it modifies to 15. CRITICAL FAILURE
I feel like it’s a bit ridiculous. A professional with expertise doing the worst they possibly can shouldn’t be the same as any random untrained person doing the worst they can.
That is why they ditched critical failures and success in tabletop D&D.
My guess is they kept it in bg3 so there would be a chance of failure on everything including the DC 2 rolls, but to be honest I don’t think that chance of failure really adds anything to the game.
I don’t understand what you mean. The game rolls automatically for lockpicking. If you roll a 1 it fails even if say the DC is 10 and you have +9 from expertise and various gear.
It’s the second shittiest common house rule, assuming you mean that if someone with a +15 bonus rolls a nat 1 on a DC 5 check, they automatically fail (possibly with a worse effect than if someone with a -1 rolled a 2).
On the other hand, there are other ways to have crit fails on skill checks that are much more palatable, like:
having a slightly worse effect when someone rolls a nat 1 and would have failed anyway
having a worse effect when someone’s total is 1 or lower
having a worse effect when rolls are failed by certain thresholds, like by 10 or more (potentially, but not necessarily, only when the roll was a nat 1)
(The worst common house rule, btw, is crit miss tables for additional effects beyond an automatic miss when you roll a 1 on an attack roll.)
If a 1 is not a fail, why do you roll at all ?
I mean if the DC is 5 and you have +15, your DM should just not make you roll (* you pass automatically). So a 1 should always be a fail.
In tabletop you shouldn’t be rolling if there is no chance of failure, although some DMs roll to see how successful the outcome is instead of just treating it as pass/fail.
The DM doesn’t necessarily have your modifiers memorized and asking what they are every time slows down play. The DM also likely doesn’t want to share the DC. The easiest fair solution is to always ask for a roll (assuming it’s possible, generically, to succeed or fail) and to then consider passes to be passes. If you only avoid asking for a roll when you know the player will make it, then you’re likely to be biased toward the players whose characters you’re more familiar with.
So a 1 should always be a fail.
RAW this is not the case. From the DMG:
Rolling a 20 or a 1 on an ability check or a saving throw doesn’t normally have any special effect. However, you can choose to take such an exceptional roll into account when adjudicating the outcome. It’s up to you to determine how this manifests in the game.
My experience with having nat 1s being auto fails and is that this results in characters who are “erratically … tragically incompetent” as well as taking away player agency (Nick Brown on rpg.stackechange explained this well). Maybe you and your players like a game like that, but I certainly don’t.
The DM doesn’t necessarily have your modifiers memorized and asking what they are every time slows down play.
Pen and paper or even a tablet exist for a reason. Having the key stats of your player characters stuck up to your GM screen or open on your second monitor is about the best use of space there is.
Besides, it doesn’t take much brain power to put together that making the rogue who’s been making short work of locks the past month roll for a simple lock under no time pressure is just silly. I get if an AL GM doesn’t know the characters but for majority of weekly groups’ GMs this is an absolute non-issue.
do you list for every character on your GM screen? Are you running games with 6+ players? And what other value does that provide? Do you also account for possible bonuses, some of which can be activated as reactions? Do you update it when values change mid-session (like if a player changes shape, enters or leaves a paladin’s aura, or attunes-deattunes an item that boosts a stat)? Because if so, that sounds like a lot of bookkeeping for very little value to me, and if not, then there’s even less value. If it’s baked into your VTT then sure, but I don’t think most groups playing in person use a VTT at the table.
What are you listing on your main screen? I’d think that all of these things would be more important than duplicating the info PCs themselves have: having a copy of the map (with everything visible or with LOS / fog of war / lighting applied to the selected PC/NPC/monster), initiative (which could be swapped out for 1-2 PC stats outside of combat), players’ passive values, everyone’s HP + status effects, notes on NPC and monster motivations, reminders (e.g., tell the Fighter they find a note from an NPC in their backpack when they say they’re getting something or during downtime), notes related to what you actually have planned for the session, etc…
Noticing that the Rogue has been doing very well on thieves tools checks and thus not making them roll to pick a lock is a clear example of the possible ways DM bias can occur, since another character (let’s say a Wizard with History expertise) with a skill bonus that’s just as high that hasn’t come up as much won’t get that same benefit. On the other hand, if you don’t have auto crit fails on 1s, then you can just let the Rogue succeed without treating the Wizard unfairly - yes, they have to roll, but there’s no chance of them failing that the Rogue doesn’t have.
That all said, if you have auto crit fails on 1s, why are you not asking for rolls all the time, anyway? Seems inconsistent, as not asking for a roll from someone who will succeed on a 1 is the same (in terms of possible outcomes) as allowing a roll but waiving the chance of an automatic failure on a nat 1. Why not just get rid of that chance in all cases?
What portion of the 18 skills 6 saves + AC 30+ tools do you list for every character on your GM screen?
Oh please, you don’t have to make this stupidly contrived. The base attributes, skills and AC are plenty enough. You might not even need the skills outside of outlandish expertise cases as even proficient skills are not that far off. You do realize the way skill scores are calculated is extremely predictable in 5e? Save DCs come from rules text most of the time, and besides, you probably should not be doing outlandish DC checks like 2 or 40 that often for it to become a trouble. When’s the last time you’ve actually needed to know the players’ tool proficiencies? You know the rogue can pick a lock and the bard can play a lute. If they have something more outlandish, they’ll let you know.
Are you running games with 6+ players?
No, because my experience tells me too large groups lead to singular players having their time in the limelight so infrequently boredom is practically assured. Even then, a table with 6 or even 10 colums is barely wider than one with 3. Like literally, go boot up Excel, paste the skill names on the left column and you’ll see very quickly that it’d all easily fit on one sheet of paper. The numbers don’t take up much space horizontally.
Noticing that the Rogue has been doing very well on thieves tools checks and thus not making them roll to pick a lock is a clear example of the possible ways DM bias can occur, since another character (let’s say a Wizard with History expertise) with a skill bonus that’s just as high that hasn’t come up as much won’t get that same benefit.
This is silly. The solution to this is to take note of the wizard’s specialties too, not to punish the rogue with having them roll pointless rolls. Your characters are not going to have that many outlandishly high scores that you couldn’t just round up these outliers and make a note of them. From all of my experience with 5e and various DMs and DMing myself, most DMs make their players roll way too many damn pointless rolls. People forget the old rule of thumb that says that you should not roll unless both success and failure are possible and both provide a meaningful outcome that carries the story forwards. If the characters are not under time pressure and they can retry endlessly, just let them have it without a roll. The rolls will feel far more suspenseful when 90% of them aren’t wasted on meaningless drudgery.
That all said, if you have auto crit fails on 1s, why are you not asking for rolls all the time, anyway?
You have it backwards. You should not be asking for rolls constantly because 1 always fails, you should only ever be rolling checks if 1 can fail! If 1 can’t fail (or 20 succeed) you just don’t roll. It’s that simple. When’s the last time you missed your mouth when trying to eat a sandwich? Doubt that’s happened to you any time recently. Similarly, don’t ask for rolls on trivial things. You don’t roll to get out of bed, you don’t roll to climb a set of stairs, you don’t roll to not choke while drinking. Accept that the player characters are good at what they do. A pro does not fail a trivial task 5% of the time, so don’t use a die to force them to. The rogue has picked locks his whole life and picking one is trivial for him, unless there is a specific circumstance that makes it otherwise, like time pressure, risk of getting caught or a particularly difficult lock.
The context of this is games where nat 1s are auto fails on skill checks, though. If your argument isn’t ultimately in defense of that, too, then I don’t have any objections to your argument (other than that it isn’t topically relevant). My objections are exclusively to the idea of not asking someone to roll because they would succeed on a 1 in a game where Nat 1s are auto-fails.
In a game where Nat 1s aren’t auto fails, then I don’t have any objections. Mechanically, both approaches - whether you do or don’t ask, and how you track to determine when to not ask - are identical, so choose the approach that makes the most sense for you, your players and your collective play styles.
Yep, that’s solid! If they got rid of the bit that the designer, Mark, clarified in the comments (“If your nat 20 isn’t a critical success, it is still a success, and if your nat 1 isn’t a critical failure, it is still a failure”) then I’d be a fan. Even with that I still appreciate the extra flexibility that it adds to the design space.
If you roll a Nat 20 or Nat 1 it raises or lowers the outcome from a crit success, success, fail, or crit fail. So Nat 20 on a roll that’s still ten below the DC takes it from a crit fail to a fail. It stops a Nat 20 on an impossible task from being a success if your skill just isn’t good enough in any way. We like playing it this way cause if our bonuses are so good (reflecting high training and skill) we can auto pass certain low level checks even on a Nat 1. It still means a Nat 1 is somewhat impactful as it stops an auto crit
[Edit]: Adding an example
Rogue attempts to pick a ‘complex’ lock with a low DC.
Complex locks require multiple successes to actually unlock, and a crit success on a pick lock check counts as 2 successes towards opening it.
Rogue has +22 to picking locks, lock DC is 10.
With a roll of 1, the result is 23, which is more than 10 above the DC meaning critical success. But since its a Nat 1 it drops to success.
So 5% chance of only getting 1 success towards the multiple required to open the lock.
Picking a lock requires an action in combat to do this, so can add a bit more intensity if the party desperately needs to open the lock quickly.
If it was outside of combat, DM would just say we unlocked it since its impossible to fail meaningfully if given enough time.
On the other end, if the a different party members bonus is +4 because they are untrained and the DC is 35, a nat 20 gives a 24 as a result which is still 10 below a success, so a crit fail. Since it was a nat 20, the result goes up from a crit fail, to just a fail meaning it mitigates the worst part of the result. And FYI, a crit fail on picking a lock breaks your lockpicks so there’s extra outcomes and narrative results gained by using the crit fail, fail, success, and crit success rules
Some people do like crits on skill checks. Other people just like rolling dice as much as possible.
The best way I’ve seen it in game was a DM making it so a natural one that you’d succeed with anyways just means you succeed in the ugliest way possible.
Like, you picked the lock, but you cut your hand on a rough edge just enough to annoy you for the rest of the day.
You made the jump, but stumbled awkwardly on landing.
That’s pretty close to the way I’ve seen it done when the DM wants a 1 to be a special number and I didn’t hate it.
A slight alteration of that is to have a successful 1 result in a complication - some of which would result in the attempted task becoming impossible or irrelevant. Maybe you pick the lock but the door is stuck or barred, or maybe you’re halfway through picking the lock when an ogre slams into it, shattering it into little more than splinters. Or you pick the lock flawlessly, but the thing you were after is missing because it was already stolen. It’s crucial that it’s something that’s out of your control with regard to the task you were performing, not that you slipped up in some way to cause the failure. It’s not perfect, but I personally like it a lot more than other implementations.
Woah, crit fail tables, ain’t nobody got time for that. I like to use crit fails as an opportunity to impose a cost or hard choice on my players, both in combat and in skill checks. But then, sometimes I just have it as a no, because it’s possible to make no mistakes and still lose.
Really, though, I always just thought that that was how it worked.
Are injury tables for hitting 0 HP are a common house rule? I can’t even remember the last time I read about a group using one, and I was subbed to multiple game tales groups on Reddit (including RPG Horror Stories).
If you’re trying to run a gritty game, I can understand wanting a player hitting 0 to have an impact. I don’t think injury tables work otherwise. And I don’t personally think 5e is a great system for gritty games overall. The rules for Lingering Injuries are in the DMG, though, so it’s not always even a house rule.
Ultimately if 1 will not fail or 20 will not succeed, why are you even rolling? While there is no default automatic success/failure rule, it’s a natural assumption that 1 and 20 are automatic fail and success based on the fact that the roll is pointless otherwise.
As someone has mentioned, “Pass” and “Fail” are not the only possible outcomes of any given roll. That’s why there are numbers on the dice besides 1 and 20.
Also, the GM doesn’t usually(and also shouldn’t, with everything else they need to keep track of) memorize every aspect of all their players’ character sheets - they don’t necessarily KNOW if the check is impossible to pass or fail.
Yeah. Larian made some really good changes to D&D, then they added crit fails to skill checks
Do you know how many times that has pissed me off? Especially on my rogue where even a 1 would have opened the damn lock.
DC 10. You roll a natural 1, it modifies to 15. CRITICAL FAILURE
I feel like it’s a bit ridiculous. A professional with expertise doing the worst they possibly can shouldn’t be the same as any random untrained person doing the worst they can.
That is why they ditched critical failures and success in tabletop D&D.
My guess is they kept it in bg3 so there would be a chance of failure on everything including the DC 2 rolls, but to be honest I don’t think that chance of failure really adds anything to the game.
That’s why I do crit fail confirms.
Yeah, as DM I’ve always house ruled that it didn’t make sense for a character to fail at the thing they’re the best at.
Though I have been known to interpret a natural 1 as a crazy external force - like an earthquake - and have them reroll at -10.
Makes it even more fun when they succeed anyway.
House Ruled? That’s RAW. Crit Fails and successes only apply to attack rolls and death saves. And that’s how it should be.
Why are you rolling in those cases?
BG3 is a video game, no DM to say “oh the rogue with a +17 doesn’t need to roll”
It was not clear from context that it was a comment about checks in BG3. I read it as a reason why they hated similar checks at the table.
No worries, we all fail perception checks every once in a while. Or spot checks, depending on edition/game.
Yes it was.
I don’t understand what you mean. The game rolls automatically for lockpicking. If you roll a 1 it fails even if say the DC is 10 and you have +9 from expertise and various gear.
Is taking 10/taking 20 not a thing in later editions?
Edit: apparently not explicitly though the dm handbook implies you should delete players auto succeed on tasks they can retry
This comment chain is about Baldur’s Gate 3.
yeah I replied to the wrong comment my b
Are those actually “crit” fails or just auto fails?
Never bothered to check if a nat one fail is any different than a nat two fail
Just auto fail. A rogue lock picking a DC10 door still has a 1/20 chance of failing the check. That’s the difference.
[nervous sweating] I’ve always run my game with crit fail skill checks. That’s normal.
Isn’t it?
Isn’t it?
It’s the second shittiest common house rule, assuming you mean that if someone with a +15 bonus rolls a nat 1 on a DC 5 check, they automatically fail (possibly with a worse effect than if someone with a -1 rolled a 2).
On the other hand, there are other ways to have crit fails on skill checks that are much more palatable, like:
(The worst common house rule, btw, is crit miss tables for additional effects beyond an automatic miss when you roll a 1 on an attack roll.)
If a 1 is not a fail, why do you roll at all ? I mean if the DC is 5 and you have +15, your DM should just not make you roll (* you pass automatically). So a 1 should always be a fail.
In tabletop you shouldn’t be rolling if there is no chance of failure, although some DMs roll to see how successful the outcome is instead of just treating it as pass/fail.
The DM doesn’t necessarily have your modifiers memorized and asking what they are every time slows down play. The DM also likely doesn’t want to share the DC. The easiest fair solution is to always ask for a roll (assuming it’s possible, generically, to succeed or fail) and to then consider passes to be passes. If you only avoid asking for a roll when you know the player will make it, then you’re likely to be biased toward the players whose characters you’re more familiar with.
RAW this is not the case. From the DMG:
My experience with having nat 1s being auto fails and is that this results in characters who are “erratically … tragically incompetent” as well as taking away player agency (Nick Brown on rpg.stackechange explained this well). Maybe you and your players like a game like that, but I certainly don’t.
Pen and paper or even a tablet exist for a reason. Having the key stats of your player characters stuck up to your GM screen or open on your second monitor is about the best use of space there is.
Besides, it doesn’t take much brain power to put together that making the rogue who’s been making short work of locks the past month roll for a simple lock under no time pressure is just silly. I get if an AL GM doesn’t know the characters but for majority of weekly groups’ GMs this is an absolute non-issue.
What portion of the
do you list for every character on your GM screen? Are you running games with 6+ players? And what other value does that provide? Do you also account for possible bonuses, some of which can be activated as reactions? Do you update it when values change mid-session (like if a player changes shape, enters or leaves a paladin’s aura, or attunes-deattunes an item that boosts a stat)? Because if so, that sounds like a lot of bookkeeping for very little value to me, and if not, then there’s even less value. If it’s baked into your VTT then sure, but I don’t think most groups playing in person use a VTT at the table.
What are you listing on your main screen? I’d think that all of these things would be more important than duplicating the info PCs themselves have: having a copy of the map (with everything visible or with LOS / fog of war / lighting applied to the selected PC/NPC/monster), initiative (which could be swapped out for 1-2 PC stats outside of combat), players’ passive values, everyone’s HP + status effects, notes on NPC and monster motivations, reminders (e.g., tell the Fighter they find a note from an NPC in their backpack when they say they’re getting something or during downtime), notes related to what you actually have planned for the session, etc…
Noticing that the Rogue has been doing very well on thieves tools checks and thus not making them roll to pick a lock is a clear example of the possible ways DM bias can occur, since another character (let’s say a Wizard with History expertise) with a skill bonus that’s just as high that hasn’t come up as much won’t get that same benefit. On the other hand, if you don’t have auto crit fails on 1s, then you can just let the Rogue succeed without treating the Wizard unfairly - yes, they have to roll, but there’s no chance of them failing that the Rogue doesn’t have.
That all said, if you have auto crit fails on 1s, why are you not asking for rolls all the time, anyway? Seems inconsistent, as not asking for a roll from someone who will succeed on a 1 is the same (in terms of possible outcomes) as allowing a roll but waiving the chance of an automatic failure on a nat 1. Why not just get rid of that chance in all cases?
Oh please, you don’t have to make this stupidly contrived. The base attributes, skills and AC are plenty enough. You might not even need the skills outside of outlandish expertise cases as even proficient skills are not that far off. You do realize the way skill scores are calculated is extremely predictable in 5e? Save DCs come from rules text most of the time, and besides, you probably should not be doing outlandish DC checks like 2 or 40 that often for it to become a trouble. When’s the last time you’ve actually needed to know the players’ tool proficiencies? You know the rogue can pick a lock and the bard can play a lute. If they have something more outlandish, they’ll let you know.
No, because my experience tells me too large groups lead to singular players having their time in the limelight so infrequently boredom is practically assured. Even then, a table with 6 or even 10 colums is barely wider than one with 3. Like literally, go boot up Excel, paste the skill names on the left column and you’ll see very quickly that it’d all easily fit on one sheet of paper. The numbers don’t take up much space horizontally.
This is silly. The solution to this is to take note of the wizard’s specialties too, not to punish the rogue with having them roll pointless rolls. Your characters are not going to have that many outlandishly high scores that you couldn’t just round up these outliers and make a note of them. From all of my experience with 5e and various DMs and DMing myself, most DMs make their players roll way too many damn pointless rolls. People forget the old rule of thumb that says that you should not roll unless both success and failure are possible and both provide a meaningful outcome that carries the story forwards. If the characters are not under time pressure and they can retry endlessly, just let them have it without a roll. The rolls will feel far more suspenseful when 90% of them aren’t wasted on meaningless drudgery.
You have it backwards. You should not be asking for rolls constantly because 1 always fails, you should only ever be rolling checks if 1 can fail! If 1 can’t fail (or 20 succeed) you just don’t roll. It’s that simple. When’s the last time you missed your mouth when trying to eat a sandwich? Doubt that’s happened to you any time recently. Similarly, don’t ask for rolls on trivial things. You don’t roll to get out of bed, you don’t roll to climb a set of stairs, you don’t roll to not choke while drinking. Accept that the player characters are good at what they do. A pro does not fail a trivial task 5% of the time, so don’t use a die to force them to. The rogue has picked locks his whole life and picking one is trivial for him, unless there is a specific circumstance that makes it otherwise, like time pressure, risk of getting caught or a particularly difficult lock.
The context of this is games where nat 1s are auto fails on skill checks, though. If your argument isn’t ultimately in defense of that, too, then I don’t have any objections to your argument (other than that it isn’t topically relevant). My objections are exclusively to the idea of not asking someone to roll because they would succeed on a 1 in a game where Nat 1s are auto-fails.
In a game where Nat 1s aren’t auto fails, then I don’t have any objections. Mechanically, both approaches - whether you do or don’t ask, and how you track to determine when to not ask - are identical, so choose the approach that makes the most sense for you, your players and your collective play styles.
If the DM doesn’t know the stat your character has the highest in and uses all the time, they have an awful memory and shouldn’t really DM.
Need to do the Pathfinder crit fails. Crit fail whenever you get 10 less than the DC, or roll a nat 1 and get less than the DC.
https://paizo.com/community/blog/v5748dyo5lkod?Critical-Hits-and-Critical-Failures
Yep, that’s solid! If they got rid of the bit that the designer, Mark, clarified in the comments (“If your nat 20 isn’t a critical success, it is still a success, and if your nat 1 isn’t a critical failure, it is still a failure”) then I’d be a fan. Even with that I still appreciate the extra flexibility that it adds to the design space.
If you roll a Nat 20 or Nat 1 it raises or lowers the outcome from a crit success, success, fail, or crit fail. So Nat 20 on a roll that’s still ten below the DC takes it from a crit fail to a fail. It stops a Nat 20 on an impossible task from being a success if your skill just isn’t good enough in any way. We like playing it this way cause if our bonuses are so good (reflecting high training and skill) we can auto pass certain low level checks even on a Nat 1. It still means a Nat 1 is somewhat impactful as it stops an auto crit
[Edit]: Adding an example Rogue attempts to pick a ‘complex’ lock with a low DC. Complex locks require multiple successes to actually unlock, and a crit success on a pick lock check counts as 2 successes towards opening it. Rogue has +22 to picking locks, lock DC is 10. With a roll of 1, the result is 23, which is more than 10 above the DC meaning critical success. But since its a Nat 1 it drops to success. So 5% chance of only getting 1 success towards the multiple required to open the lock. Picking a lock requires an action in combat to do this, so can add a bit more intensity if the party desperately needs to open the lock quickly. If it was outside of combat, DM would just say we unlocked it since its impossible to fail meaningfully if given enough time.
On the other end, if the a different party members bonus is +4 because they are untrained and the DC is 35, a nat 20 gives a 24 as a result which is still 10 below a success, so a crit fail. Since it was a nat 20, the result goes up from a crit fail, to just a fail meaning it mitigates the worst part of the result. And FYI, a crit fail on picking a lock breaks your lockpicks so there’s extra outcomes and narrative results gained by using the crit fail, fail, success, and crit success rules
Oh cool! That makes sense and addresses my concerns. I really like the sound of that whole system, overall.
Is what you described a house rule or RAW in PF2E?
It’s RAW ☺️
https://2e.aonprd.com/Rules.aspx?ID=319
Crit fails like that are referenced as an optional rule in the DMG, minus the Nat 1 part
Some people do like crits on skill checks. Other people just like rolling dice as much as possible.
The best way I’ve seen it in game was a DM making it so a natural one that you’d succeed with anyways just means you succeed in the ugliest way possible.
Like, you picked the lock, but you cut your hand on a rough edge just enough to annoy you for the rest of the day.
You made the jump, but stumbled awkwardly on landing.
Etc etc
That’s pretty close to the way I’ve seen it done when the DM wants a 1 to be a special number and I didn’t hate it.
A slight alteration of that is to have a successful 1 result in a complication - some of which would result in the attempted task becoming impossible or irrelevant. Maybe you pick the lock but the door is stuck or barred, or maybe you’re halfway through picking the lock when an ogre slams into it, shattering it into little more than splinters. Or you pick the lock flawlessly, but the thing you were after is missing because it was already stolen. It’s crucial that it’s something that’s out of your control with regard to the task you were performing, not that you slipped up in some way to cause the failure. It’s not perfect, but I personally like it a lot more than other implementations.
Woah, crit fail tables, ain’t nobody got time for that. I like to use crit fails as an opportunity to impose a cost or hard choice on my players, both in combat and in skill checks. But then, sometimes I just have it as a no, because it’s possible to make no mistakes and still lose.
Really, though, I always just thought that that was how it worked.
That’s not as bad as injury tables for going down
Are injury tables for hitting 0 HP are a common house rule? I can’t even remember the last time I read about a group using one, and I was subbed to multiple game tales groups on Reddit (including RPG Horror Stories).
If you’re trying to run a gritty game, I can understand wanting a player hitting 0 to have an impact. I don’t think injury tables work otherwise. And I don’t personally think 5e is a great system for gritty games overall. The rules for Lingering Injuries are in the DMG, though, so it’s not always even a house rule.
Crit fails on skill checks have been houseruled into the game for ages, this is not something cooked up by Larian
Crit fail and success for skill checks is a variant rule in the dmg (maybe even discussed in the PHB)
Ultimately if 1 will not fail or 20 will not succeed, why are you even rolling? While there is no default automatic success/failure rule, it’s a natural assumption that 1 and 20 are automatic fail and success based on the fact that the roll is pointless otherwise.
Tiered DC. I don’t know if 5th has that as an official rule or it’s a common house rule.
Rolling to seduce a dragon? Nat 20: he/she laughs at you, you fail. Nat 2: does a 22 hit your AC?
It’s mentioned somewhere in Chapter 8 of the DMG.
As someone has mentioned, “Pass” and “Fail” are not the only possible outcomes of any given roll. That’s why there are numbers on the dice besides 1 and 20.
Also, the GM doesn’t usually(and also shouldn’t, with everything else they need to keep track of) memorize every aspect of all their players’ character sheets - they don’t necessarily KNOW if the check is impossible to pass or fail.
Well yes, but actually no. BG3 crit fail and DMG crit fail are different