The d100 illusion
Percentile dice look scientific. A hundred steps of granularity are not the same thing as a hundred meaningful differences.
01Precision is not resolution
A d100 reads like a probability statement you could put in a research paper. Sixty-three percent. Forty-seven percent. The number feels engineered rather than guessed, and that impression is part of what sold percentile systems — Call of Cthulhu, Rolemaster's skill rolls, RuneQuest's characteristic-derived chances — to players who wanted their odds legible. You can see exactly where you stand. The die obliges by producing exactly the number you named.
What the die is actually doing
- 01Granularity — the count of distinct values a die can produce (d100 = 100)
- 02Resolution — how many of those values produce a meaningfully different outcome at the table
- 03Live range — the band of skill values where a roll is genuinely uncertain; in most percentile systems, roughly 40–80%
- 04Interface function — a number's ability to be emotionally parsed and compared by players, independent of its mechanical precision
The problem is that granularity and resolution are not the same thing. Granularity means the die can output a hundred distinct values. Resolution means those values correspond to something different in the fiction. A skill sitting at 47% and one sitting at 52% are five whole points apart on a hundred-point scale, which sounds meaningful — but across a session those two skills will produce distinguishable results only very rarely. The standard deviation of a pass/fail roll at fifty percent is large enough that you would need hundreds of rolls before the difference between 47 and 52 became statistically visible. The precision is real. The meaningful difference is mostly imaginary.
This is the d100 illusion: because the number is specific, it feels like it was calibrated. In practice, most percentile systems cluster usable skills between roughly 40 and 80. Below 40, failure is so common the skill barely functions. Above 80, success is so routine the roll is a formality. The live range — the stretch where rolling actually creates uncertainty — is perhaps thirty or forty points wide, no more granular in practice than a target number on a twenty-sided die. The other sixty points exist on the sheet without doing much mechanical work.
Many percentile games improve skills by small increments — two points here, three there — which feels like fine-grained reward.
Advancement sharpens the problem. Many percentile games improve skills by small increments — two points here, three there — which feels like fine-grained reward. But a three-point gain on a skill already sitting at 65% changes your success chance by three percentage points. Your expected outcome across a normal session changes by far less than one roll. The reward is legible without being felt.
None of this is a flaw that breaks a game. Call of Cthulhu has run smoothly for decades on exactly this chassis. What it means is that the hundred-step scale is doing a different job than pure mechanical precision: it is doing interface design, giving players a number they can emotionally parse and compare. That is a real function. Just do not confuse the readability of the number with the resolution of the die. The d100 is granular. It is not, for most of that range, particularly fine.