The hero takes a running start before a chasm, the game master holds a twenty-sided die, and a pause hangs over the table: what Difficulty Class (DC) should the hero face, 10 or 15? Miss in one direction, and in an hour the party's journey will turn into a parade of easy victories, where no one even holds their breath before a roll anymore. Miss in the other, and players will start getting angry at a series of frustrating failures that make their characters look helpless. The success threshold determines what each attempt is worth, how fast the scene moves, and what the people at the table feel. That's why we've compiled a working scale of thresholds with reference numbers, clear criteria for choosing them based on the scene and the hero, and the common mistakes that make the dice stop mattering to anyone.

What a d20 Success Threshold Actually Represents

Informative infographic illustration: a row of twenty facets or steps, illuminated with a gradient from green to red, with thresholds marked 5, 10, 15, 20. Clean editorial style.

A twenty-sided die has equal faces, and each carries exactly a 5% weight. Hence a simple formula: to find out the chance of success, subtract the threshold from 21 and multiply by five. A DC of 10 gives 11 winning faces (from 10 to 20), which is about 55%. At a DC of 15, there are six winning faces, which is 30%. A DC of 20 leaves a single face and a 5% chance. We aren't counting modifiers here yet.

It's useful to keep the numbers in mind, but something else is more important at the table: each number feels different. The DC translates probability into the language of the scene. The player doesn't calculate percentages; they feel how scary it is for the hero to jump from roof to roof.

A character's modifier shifts the whole picture. Every bonus point adds one winning face, meaning a +3 to the roll adds 15 percentage points. A rogue with +3 against a DC 15 lock is effectively rolling against 12, and their chance grows from 30% to 45%. That's why the same lock looks different to an experienced thief than it does to a village blacksmith, and you should choose the threshold keeping in mind who is rolling. How bonuses fit into the mechanics overall is detailed in our guide on d20 skill checks.

The main thing a d20 roll establishes is a sense of risk. The DC answers the question "how dangerous is this for the hero right now?" and hints to the player whether it's worth risking. A low number says: act boldly. A high number warns: think about whether there's a workaround. A well-chosen DC makes the player hesitate before rolling, and it's for the sake of that hesitation that the dice are picked up at all.

The Difficulty Scale: From "Almost Certain" to "Near Miracle"

Illustration of a staircase or mountain path with five marks of varying difficulty: from an easy path at the bottom to a steep cliff with a tiny figure of a hero near the top. A metaphor for increasing difficulty.

It's easiest to keep five reference points in mind: that way you won't have to leaf through a table in the middle of a scene. Intermediate numbers are easy to slot between them if the scene needs nuance. Below are examples from three genres, so the scale isn't tied just to swords and dragons.

Now let's calculate the odds. To beat a DC of 17, you need to roll a 17, 18, 19, or 20, which is just 20%. At 20, you're left with 5%. If such thresholds come one after another, the player fails repeatedly and stops planning altogether: bold moves don't work anyway, so why invent them? The top of the scale is good as a rare spice, for moments when the hero deliberately asks for trouble and knows it.

Working rule: the majority of checks in a session should lie between 8 and 14. That's a chance from 65% to 35%, where the outcome isn't predetermined, but success doesn't become a mere formality. Fives are fine for warm-ups and for actions where the hero is a master, and leave 17+ for the climax. If you want to double-check this math for specific characters, look into our separate breakdown: d20 skill checks.

A final question: should you state the DC out loud? An open DC gives the player a conscious choice. Seeing a 16 for a jump and a 9 to edge along the ledge, they weigh it and decide for themselves. Tactical forks work best this way. A blind DC is appropriate where the number itself gives away information: if the attempt to catch an NPC in a lie requires an 18, the player immediately realizes a professional is lying to them. Alternate both options, and skill checks will stop being predictable.

How to Adjust the DC for the Scene and Character

The number on the scale describes the hero's position at a specific moment. So first, look at the scene: what's around, who's interfering, how much time is left. The character comes next.

What shifts the threshold up or down:

Let's take the same deception of a guard at the gates. A charismatic diplomat with forged papers in their pocket will manage on an 8–10; for them, it's a routine conversation. A clumsy rookie who blushes and mixes up their lord's name will get a 14. And if the guard just caught another liar, both DCs will rise: the situation just got tougher for everyone.

Sometimes a check isn't needed at all. Walking down the street, asking a barkeep a question, opening an unlocked door: there's no risk, so no need to roll. Sometimes the opposite is true, when an action is fundamentally impossible. A hero without wings won't take flight off a roof, and no number will change that, so it's more honest to reject the attempt and explain why.

And the final filter before a roll: what happens on a failure? If the answer sounds like "nothing, try again," the roll is just wasting time. A good skill check in a text RPG puts something on the line: an alarm, a wasted minute, ruined trust. When failure threatens nothing, let the action simply succeed, and think about the check's difficulty only when failure actually pivots the plot.

Critical Failure and Critical Success: What a 1 and a 20 Should Mean

Diptych: on the left, a d20 showing a one with a broken object on the table (a cracked lantern, scattered papers), on the right, a d20 showing a twenty with a golden glow, a hero finding a rare artifact. Contrast of two extremes.

A natural 1 and a natural 20 happen on average once every twenty rolls, which is why they are remembered. If a one just results in the hero stumbling and everyone having a laugh, players quickly stop fearing it. The cost must be felt in the scene: a wound that hinders them later, a broken or confiscated item, a raised alarm, a soured relationship with a key contact, a closed off opportunity.

It's a similar story with a twenty. The player would have gotten "it worked" with a normal success anyway. A natural 20 deserves something extra that will come in handy later: an additional clue, a debt of gratitude from an NPC, extra time on the clock, or a piece of loot.

Three examples of varying severity for the exact same scene involving breaking into a warehouse:

A natural 20 in the same scene: the hero not only picks the lock but also finds shipping manifests with the client's name right there on the desk.

Controversial Homebrew: Do Crits Work on Any DC?

A common rule is "a 1 is always a failure, a 20 is always a success" regardless of the DC. With this, a hero with a massive bonus fails the simplest action 5% of the time, and a rookie does the impossible 5% of the time. A DC of 5 suddenly carries a 5% chance of catastrophe, and mundane checks start getting scary. A cleaner approach works better: don't call for a roll at all where there's no risk, but where there is risk, a 1 and a 20 decide the outcome independently of difficulty and modifiers. Then a critical outcome remains a rare event and doesn't turn into a lottery at every step. How this meshes with other rules is something we covered in our text on d20 skill checks.

Finding the Balance

The guideline is simple: the consequence of a critical failure should complicate the path while leaving it open. The death of a hero or a permanently lost plot key because of one bad die roll kills engagement faster than any boredom. Match the severity to the stakes of the scene. Incidentally, an equally weighty reward for a critical success helps maintain balance: the player hopes for a twenty just as intensely as they dread a one.

What to Do After a Failure: How Not to Turn a Miss into a Dead End

A hero at a locked door that didn't yield; nearby another path opens up – a narrow passage or window in the wall. Conveys the idea of 'one door closes, but the story moves on'.

The most common GM failure happens right after the roll. The hero failed to pick the lock, and the scene halts: the door is closed, the player stares at it, tries again, then again. Five minutes later, the die has become a zero-stakes lottery, and interest fades. This is cured by one rule: failure drives the story forward.

Failure must change the world. While the hero fiddled with the lock, time passed, and the guard changed. The lockpick broke, and the quiet route is closed. Someone in the hall heard a scratching sound. The player is still at the same door, but the situation around them is different, and they have to solve a new problem.

A separate issue is rolling for the exact same thing until you win. If a retry is free, the DC means nothing: sooner or later, the d20 will show the right number. Therefore, every subsequent attempt must cost more. Raise the difficulty by a couple of points, consume a resource (the torch burns down, ammo runs out), or outright say this opportunity won't present itself a second time.

Three techniques that save almost any scene:

If nothing changes after a failure, players quickly stop expecting anything from the dice: roll or don't roll, it turns out the same. In that case, checks should be called less often and only when something is truly at stake. More on when a roll is even appropriate can be read separately.

A number on a die face means little on its own. The thrill appears when the player knows that a 1 will be followed by something tangible, and a success will open up things that wouldn't have happened otherwise.

Common DC Mistakes and How to Fix Them

Almost all DC-related mishaps boil down to six mistakes. Run through this list before your next session: chances are, you'll recognize yourself in at least one point.

In a text RPG with an AI storyteller, the machine makes these same decisions. In AI Quest, under the "AI Quest Rules" mode, the narrator dynamically assigns a difficulty from 1 to 20 tailored to the specific scene and hero. The server rolls before the player even sees the number, so tweaking the DC after the die drops is impossible. And why an AI without such a framework eagerly agrees to any prompt is broken down in our article on how dice bring stakes back to the story.

What DC is considered "normal" on a d20?

The benchmark is 10. Without modifiers, a 20-sided die rolls a 10 or higher 55% of the time, so a hero with no special advantages passes such a check slightly more often than they fail. It's a handy starting point: use 5 when failure would require sheer clumsiness, 14 where genuine skill is mandatory, and save 17 and above for actions on the very edge of possibility.

Should a 20 be an auto-success and a 1 an auto-failure regardless of the DC?

Yes, and it's a useful safety valve. The chance at each extreme is 5%: the master occasionally slips up, the rookie occasionally gets lucky, and this keeps both on their toes. One caveat regarding the natural 20: if an action is fundamentally impossible, like jumping to the moon, don't allow a roll in the first place. A natural 20 saves the very hard, but the utterly impossible shouldn't even make it into the rolling phase.

How often should I ask for a roll in a session?

A stopwatch won't help here; look at the scene. A roll is needed where there's risk or resistance and where failure changes something. Looking around, chatting up a tavern keeper, or opening an unlocked door can be done without dice. Before every roll, ask yourself what happens on a failure. If the answer is "nothing," the die is superfluous. Where the boundary lies between a roll and a mundane action is discussed in our separate breakdown on skill checks.

What if the player fails a vital check and the plot gets stuck?

Let the failure drive the story forward. Time has passed, guards are alerted, the old route is blocked, but a new–albeit costlier–detour has appeared. Never let them spam the exact same check for free: make every retry harder. And in the future, don't hide vital plot clues behind a single die roll. If the game cannot proceed without it, ensure the hero gets at least part of it without needing to roll.

Can I change the DC after the roll?

It's better not to. If you nudge the difficulty after the number has fallen, a failure instantly looks like a GM's whim, and the player loses faith in the dice. Lock the DC before the roll. If circumstances change right before it, recalculate and then roll. Realize you messed up the number after the fact? Leave the result as is and adjust your bar for future checks.