Armour break runs all fight. Break only pays in the window
✍️ Alice・愛麗沙
These two do not live in the same bucket, and they are more than three times apart in value. Armour break sits in the second sub-multiplier of the damage-increase bucket — the same slot as the “damage taken” debuff an ally applies — and it is live for the entire fight. The break stat sits in the defenceless bucket, and it only wakes up during the few seconds an enemy is inside the defenceless window. In a three-minute boss fight with four breaks of eight seconds each, that window is 17.8% of the clock: +1,000 break stat is worth +15.9% inside the window, which multiplies down to +2.8% across the fight. The same slot carrying +10% armour break is +10% across the whole fight. Unless your party can hold the window open long and often, the break stat ranks far below where its printed number suggests.
Two equations, two addresses
Armour break — the second layer of damage-increase bucket C:
C = (skill power and every damage-increase source, summed) × (1 + armour break% + ally damage-taken increase%)
The break stat — defenceless bucket G:
G = (1 + break stat ÷ 5250) × (1 + break damage%) + flat term, with the whole term multiplied by 1.5 while break is stacked to its limit
The difference is not in the numbers. It is in the clause attached to the second one: only while the enemy has been broken and is inside the defenceless window. When the window is shut, G = 1. Every point of break stat you own is, for those seconds, not there.
Which break states open which damage types is settled — the state names and their pairings are confirmed, even though the percentages inside the cells are not:
| Break state | Smash | Multi-hit | Element gaps |
|---|---|---|---|
| 擊飛 | +10% | +10% | 火焰 +5%、電氣 +5% |
| 冰凍 | +10% | +10% | 寒氣 +5%、電氣 +5% |
| 魅惑 | +10% | +10% | 精神 +5%、神聖 +5%、暗黑 +5% |
| 暈眩 | +10% | +10% | 火焰 +5%、寒氣 +5% |
| 精神 | +10% | +10% | 毒 +5%、出血 +5%、精神 +5% |
| 石化 | +10% | +10% | 神聖 +5%、暗黑 +5% |
The six break states and the damage types each one opens are confirmed by the TW official game guide 3474880 on Break (source text pulled 2026-09-01): it lists knockback, stun, mind, charm, freeze and petrify, each pairing "defenceless, smash, multi-hit" with two or three elemental or affliction damage increases, and our matrix matches that text exactly. But the official text says only "damage increases" — it publishes no percentage. The +10% / +5% in each cell still comes from the KR community damage formula (the maintained DC edition), medium confidence, with the application method never officially confirmed.
Multiply by the window share
Take a 180-second boss fight in which the party lands four breaks, eight seconds each — 32 seconds, or 17.8% of the fight.
Inside the window, with +1,000 break stat:
- Before: G = 1 + flat term 0.20 = 1.20
- After: G = 1.1905 + 0.20 = 1.3905
- Gain inside the window: 1.3905 ÷ 1.20 = +15.9%
Across the whole fight: 15.9% × 17.8% = +2.8%
The same slot carrying +10% armour break, assuming nothing else is sharing that sub-multiplier, is ×1.10 for the whole fight = +10%.
2.8 against 10. This is not a question of which one is slightly better.
That 0.20 needs an asterisk. Two Korean sources disagree here: one writes the flat term as +0.20, the other hardcodes +0.60 (the second may already have folded tag damage into it). We record both and pick neither. Redo the arithmetic with 0.60 and you get 1.7905 ÷ 1.60 = +11.9%, which times the window share is +2.1%. The two versions point the same way — whichever number is right, the break stat loses.
What would flip the break stat
A higher window share. That is effectively the only variable. Replace 17.8% with 40% — a party with strong break tools, long windows and a boss with enough health to keep the fight going — and +2.8% becomes +6.4%, which starts to approach armour break.
The stacked-to-the-limit break state. For its duration the whole G term is multiplied by 1.5, so the break stat’s absolute contribution scales with it. A party that produces that state reliably is doing different arithmetic from the one above.
Being the class that does the breaking. At that point the break stat is not a damage option on your sheet, it is your function in the group.
The official guide put two limits on break extension, and parties still trip over them
The official guide page on break carries two sentences that bear directly on how you build a party:
The break-extend effect can only be applied once per player while a monster is broken.
When several players use break-extend skills on the same monster, the extended break duration shrinks progressively, and no monster is extended beyond a specified cap.
On the Taiwanese server the guide names warrior and frost mage as the classes holding a break-extend skill. The Korean server has a third, the knight class that Taiwan has not received.
Translated into party language: the second person carrying break extension is worth a great deal less, and the third is worth close to nothing. This is not community inference — it is mechanism written into the official text. A party that wants a higher window share should be raising the number of breaks, not the headcount of extenders.
One consequence follows immediately, and it explains a tier-list result that otherwise looks odd. The frost mage’s place in team dungeons rests on holding this function, not on its damage ceiling — and because the function slot is singular, the holder is hard to displace and impossible to duplicate usefully.
Four names, disentangled once
- Break — the action of pushing an enemy into the defenceless state.
- The break stat — a number inside the defenceless bucket. Live only inside the window.
- Armour break — a percentage in the second layer of the damage-increase bucket. Live all fight.
- Defence reduction — feeds the denominator of defence mitigation. Live all fight, and each point is worth more than the last as you stack it.
There is a second pair that gets confused just as often: rage and counter. Rage (red aura) is fragile to smash-judged attacks and takes less from multi-hit-judged ones; counter (blue aura) is the exact mirror. That is the boss changing which damage judgement it fears — a different mechanism from the break window, and mixing the two into one calculation produces nonsense. The “smash enhancement / multi-hit enhancement” printed on runes is a third thing again, and the trap is the same shape: same words, different system.
What to do with this
- Solo, or content with short windows: prioritise the always-on multipliers — armour break, defence reduction — and put the break stat behind them.
- Fixed party that opens windows reliably: measure your own window share first. Four breaks of eight seconds and eight breaks of ten seconds differ by a factor of two, and a factor of two flips the ordering.
- One break-extender per party is enough. The second one should spend that skill slot on something else.
- When you see the word “break”, ask which of the four it is. Four terms separated by more than a factor of three, discussed as if they were one, wastes everybody’s time.
What could overturn this. The 5250 divisor, and the rule that G is live only inside the window, come from the long-maintained Korean community damage formula, and that section is flagged medium confidence — lower than most of the chain. The flat term is a straight contradiction between two sources, 0.20 against 0.60, and we print both. The two break-extension limits are official text and carry the highest confidence here. The 17.8% above is an illustrative assumption — if a fixed party measures a sustained window share above thirty percent, the break stat’s ranking should move up and this page’s conclusion moves with it.
