What a Difficulty Number Says
A target number is not a threshold — it is a probability. Read it that way, and most published numbers turn out to be making claims nobody intended.
01The Number Is a Bet
Every target number in a resolution system is, underneath the language of difficulty, a statement about frequency. Set DC 15 in a system where a trained character rolls 1d20 plus five: that character hits the number fifty-five percent of the time. Set it at DC 20 and you've just said this thing happens thirty percent of the time. Neither choice is neutral. Both are claims about how often the fiction should produce that outcome — and the designer who wrote the number usually thought about "hard" or "heroic" rather than sixty or thirty.
That gap between intuition and arithmetic is where most difficulty systems go quietly wrong.
The core translation
- 01target number — a threshold expressed as a difficulty label; also, implicitly, a probability statement about outcome frequency
- 02flat die — a die where every result is equally likely; moving the target by one point shifts the success chance by one equal step
- 03median build — the character a typical player will actually have at the relevant point in play; the right benchmark for calibrating difficulty, not the maximum or minimum
The mechanism is straightforward. Any flat die — a single uniform die — divides the number line into pass and fail at exactly the point the target sits. Move the target one step and you move success probability by exactly one step, because every face is equally likely. So the target number and the probability are the same claim stated in two different languages, and it helps to be fluent in both.
02What "Hard" Actually Means Across Systems
The trouble is that different systems put the same English words on wildly different probabilities. "Hard" in one game might sit at a fifty percent chance for a skilled character; "hard" in another might sit at twenty. Neither label is meaningless, but neither is interchangeable, because the die behind it is different.
A d20 system gives each pip of modifier or difficulty exactly five percent of weight. A percentile system gives each pip exactly one percent. A pool of dice gives each additional die a curved, context-dependent contribution — shifting the difficulty target by one is not the same as shifting it by one on a flat die, because you're moving a threshold across an uneven distribution. Comparing a "DC 15" to a "Difficulty 3" to a "target 55%" requires translation, and translation requires knowing the full range of the die and the realistic modifier the typical character brings.
That number is what the designer actually decided, whether they knew it or not.
This is why the same competence word — "skilled," "professional," "trained" — produces such different tables. A designer writing "skilled characters should usually succeed at moderate tasks" can encode that as anywhere between fifty-five and eighty percent depending on how they set the baseline modifier. At sixty percent the skilled character fails two sessions in five. At eighty percent failure is memorable rather than routine. Those are different games, and neither is wrong, but neither is what "usually" naively suggests.
The design stress-test
- 01Pick any difficulty label in a system ("hard," "moderate," "heroic," etc.)
- 02Find the realistic modifier for a character the label is aimed at
- 03Calculate the hit percentage
- 04Ask whether that percentage matches what the label implies — this is the gap where most published systems leak
03Reading a Published Number Backwards
The useful practice is to read any target number backwards from the character sheet outward. Take the realistic modifier a player character will have at the point in play the number is intended for — not the maximum, not the minimum, the median build — and calculate the hit percentage. That number is what the designer actually decided, whether they knew it or not.
Do this across a stat block or a skill list and patterns emerge. Many systems have a cluster of "moderate" difficulty numbers that land between forty-five and fifty-five percent for their median character — close enough to a coin flip that the word "moderate" starts to feel ironic. Others have "easy" difficulties that a well-built character hits ninety percent of the time, which means the roll exists almost entirely to catch fumbles rather than to generate genuine uncertainty.
Neither outcome is catastrophic if it's intentional. A roll that mainly catches fumbles can still serve a dramatic function. A near-coin-flip at moderate difficulty produces a specific texture of play — high variance, frequent reversal — that some groups love. The problem is only when the probability and the intent are strangers to each other, when the referee calls for a roll expecting the rogue to "probably make it" and the math says forty-two percent.
Once you read difficulty numbers as probability statements, you can calibrate rather than guess. You know what you're asking for when you set or raise a target number: not a feeling of hardness, but a specific fraction of attempts that will fail, across all the moments in all the sessions where that number gets rolled against. That is what the number says. It is worth listening to.