Does crit rate have a ceiling? 50%, and you never reach it
✍️ Alice・愛麗沙
The crit stat converts to a base crit rate with a closed form: 50% × stat ÷ (stat + 2000). It is a curve that flattens as it climbs and approaches 50% without ever arriving — 2,000 points buys 25%, 6,000 points only reaches 37.5%, and even 12,000 gets you 42.86%. So the familiar advice that “3,000 crit is enough, 4,000 is plenty” is not a mechanical wall; it is an economic crossover on a perfectly smooth curve, the point past which the same resources buy more damage as smash, multi-hit or skill power. The only way to push crit rate past half is a different kind of source: the flat “crit rate +X%” printed on runes, passives and artefacts, which is added on top and is not bound by the 50% asymptote at all. The crit multiplier has no ceiling of any kind. These constants are Korean-server community measurements; Taiwan’s Season 0 runs the same structure, but no official figures exist for either server.
One expression generates the entire table
Start with the skeleton. Crit is an independent judgement multiplier in the damage chain: the hit rolls once, and if it lands the damage is multiplied by the crit multiplier. What that multiplier is worth splits into two halves — the rate and the multiplier — each with its own expression.
The rate half is tidy enough to memorise:
| Crit stat | Base crit rate | Crit multiplier |
|---|---|---|
| 1000 | 16.67% | ×1.60 |
| 2000 | 25.00% | ×1.80 |
| 3000 | 30.00% | ×2.00 |
| 4000 | 33.33% | ×2.20 |
| 5000 | 35.71% | ×2.40 |
| 6000 | 37.50% | ×2.60 |
| 8000 | 40.00% | ×3.00 |
| 10000 | 41.67% | ×3.40 |
| 12000 | 42.86% | ×3.80 |
Base crit rate = 50% × stat ÷ (stat + 2000) — it approaches 50% and never reaches it. Crit multiplier = 1.4 + stat ÷ 5000. Both are closed forms of the KR community formula; our own recomputation matches their measured tables exactly. Only flat crit-rate percentages (runes, passives) push the rate past the 50% ceiling.
Drawn as a line, the part that matters is unmistakable:
- Crit rate
The curve is steep at the bottom. Going from zero to 2,000 buys a full 25 percentage points of crit rate, which is superb value. Past the middle it starts hugging an invisible line at 50%, and everything you add after that only nudges digits behind the decimal point.
This is not a threshold someone put in the game. It is the expression’s own limit. In 50% × stat ÷ (stat + 2000), once the stat is large enough that 2,000 is negligible beside it, the fraction tends to 1 and the whole term tends to 50% × 1. Always a little short, forever.
The truth about “3,000 is enough”: that is a price crossover, not a wall
Korean community shorthand grades the stat: 3,000 crit is “efficient”, 4,000 is “semi-efficient”, anything past 5,000 is “inefficient”. The grading is fine. What goes wrong is the reading — people take it to mean crit is capped at 3,000 and you should stop. That is a misreading.
What it actually means: the crit-rate curve starts going flat around 3,000, so from there each additional point of crit returns less final damage than the same point spent on smash or skill power. That is two diminishing curves crossing, and the crossing point moves with your whole build. It is not a fixed number. In the same Korean thread, the author’s own recommendation is to “wrap up at 2,500 to 3,000 and put the rest into skill power, smash and extra hit” — note that he says wrap up, not forbidden zone.
Put differently: “how much crit should I stack” has no universal answer, only “on your current sheet, does the next point of crit still beat the next point of smash”. To see that, feed your numbers to the stat efficiency calculator and let it recompute against your own buckets. It will beat memorising any table, including the one above. If you want to compare whole loadouts rather than single stats, that is the builder.
The crit multiplier: the other half, slow to grow and unbounded
Crit rate hits a wall. Crit multiplier does not. Its expression is 1.4 + stat ÷ 5000: with zero crit you already multiply by 1.4, every 1,000 points adds 0.2, and the line goes up forever with no asymptote in sight.
Put the two halves together and something unreasonably clean falls out. With no external crit-rate percentage on your character, the entire crit multiplier region collapses to:
F = 1 + crit ÷ 10000
The algebra is worth seeing once, because it explains why the collapse is exact rather than approximate. The expected-value form is 1 + rate × (multiplier − 1), which is 1 + [0.5c/(c+2000)] × (0.4 + c/5000). The second bracket rewrites as (c + 2000)/5000 — and that factor is precisely the denominator of the first. They cancel, leaving 0.5c/5000, which is c/10000.
We recomputed it here rather than taking it on trust, and it reproduces an independent measured table with zero error: 3,000 → ×1.3, 4,000 → ×1.4, 5,000 → ×1.5, 6,000 → ×1.6. That is the strongest cross-check in the whole damage model — one source hands over a formula, an unrelated source hands over measurements, and they agree to the last digit.
The practical corollary matters more than the elegance. If you are wearing crit-rate% or crit-damage% runes, the collapse no longer holds. The added percentage lifts the rate half off the 50% asymptote, which means the crit stat itself becomes worth more than the tidy formula says. Crit is not an isolated number; it compounds with your runes. Which effects reference which judgement, and why a rune’s “smash” is not your skill’s “smash”, is its own article.
The variables that would overturn everything above
A clean model still cannot paper over two cracks in reality, so here they are stated plainly.
Runes rewrite the priority order. Carrying crit-rate or crit-damage runes raises the marginal value of the crit stat sharply. A Korean-server measurement has 3,000 crit under a particular two-rune combination moving the average damage gain from single digits to over thirty percent. So “how much crit should I stack” can never be answered in isolation from “what runes are you wearing”.
Monsters have crit resistance, and it is not in the formula. High-end content subtracts a slice straight off your crit rate. One hard dungeon has been estimated at around −25%, and one greatsword test simply assumed 30% resistance, which halved crit’s marginal value outright. The existence of this mechanic is fairly well established; the actual numbers are not — community measurements scatter badly and we rate their confidence low. The practical consequence is what counts: in content that strips crit rate, skill power, smash and multi-hit overtake crit. Which is exactly why “crit first” depends on what you are fighting and cannot be treated as a law.
One sentence to close on: crit rate has a ceiling, the crit multiplier does not, and the real priority order is decided jointly by your runes and by the monster in front of you. Any guide that reduces it to a single fixed number has left out all three.
What would overturn this page: the closed form. If anyone produces a training-dummy series where the observed crit multiplier region diverges from 1 + crit/10000 on a character carrying no flat crit-rate or crit-damage sources, the derivation above is wrong and this article gets rewritten with the correction dated. The softer claim — that monster crit resistance exists — would fall to a measured run showing identical crit frequency inside and outside high-end content.