Where should the next 100 points go?
Enter your six enhancement stats. The page ranks what another 100 points would actually buy, highest first.
Need your whole gear sheet in the answer? Use Stat Efficiency. This page compares six enhancement stats only, and is faster for that. →
Your current stats
The 2,000 default is a round number to make the table readable. Replace it with your own.
What another 100 points buys
Best right now: Skill power (+0.952%)
| # | Stat | Bucket | Divisor | Damage gain |
|---|---|---|---|---|
| 1 | Skill power 스킬 위력 F03 · MEASURED · confidence HIGH (math) / MED (label) ⚠ Sits in a different bucket (C), so it does not dilute the five above. The source writes the stat only as 스위, most likely skill power, but one secondary source reads it as something else: the maths is high confidence, the label is not. | C damage increase | 8,500 | +0.952% |
| =2 | Multi-hit enhancement 연타 강화 F05 · MEASURED · confidence HIGH | D enhancement | 8,500 | +0.574% |
| =2 | Smash enhancement 강타 강화 F06 · MEASURED · confidence HIGH | D enhancement | 8,500 | +0.574% |
| =2 | Area enhancement 광역 강화 F07 · MEASURED · confidence MED-HIGH | D enhancement | 8,500 | +0.574% |
| 5 | Ultimate enhancement 궁극기 강화 F09 · MEASURED · confidence MED ⚠ Divides by 8750, not 8500 — and this row is only MED confidence. | D enhancement | 8,750 | +0.558% |
| 6 | Combo enhancement 콤보 강화 F08 · MEASURED · confidence MED-HIGH ⚠ Modelled at full combo (100+). Below that it scales n/4: 1/4 from 1 combo, 2/4 from 10, 3/4 from 50. | D enhancement | 17,500 | +0.279% |
Percentages are relative gains in final damage. Per-skill base coefficients are unpublished, but they cancel out of any comparison — so the ranking is exact even though an absolute damage number is not obtainable.
Why spreading your stats evenly achieves nothing
Multi-hit enhancement / Smash enhancement / Area enhancement all divide by 8500, and inside their bucket they are added together. Putting 100 into any of them moves that sum by the same 100/8500 — so which one is already high makes no difference at all.
What does separate them is the divisor: Combo enhancement divides by 17,500, worth roughly half as much; Ultimate enhancement divides by 8,750, worth slightly less.
So "spread it evenly" and "stack one" are the same decision as far as damage is concerned. The real differences are elsewhere: how much of your class's skill mix is multi-hit versus smash, and which type your runes and engravings give a damage % to.
This was measured rather than reasoned (F26): training dummy, 5+ cycles per configuration, stats swapped with potions while the sum was held at 5,112. Swapping smash and multi-hit produced damage identical down to the ones digit.
How the model works
- Inside a bucket values add; between buckets they multiply. It is the single most important structural fact in the whole calculation.
- The five enhancement types — multi-hit, smash, area, combo, ultimate — are additive with each other and multiplicative only inside themselves. Treating them as multiplicative with each other is the most common error in community calculators, and it makes stacking one stat look far better than it is.
- Smash and multi-hit share one pool, and only their sum matters. Measured on the Korean server, superseding an older belief that smash used /85 and multi-hit /90.
- Skill power sits in bucket C and multiplies the D bucket. So the fuller your D bucket gets, the less another D point is worth and the better skill power looks by comparison. The ranking reflects that on its own.
- Each point is worth proportionally less as the stat grows, which is why a fixed "best stat" tier list is wrong for most readers, and why this asks for your current values instead.
The formulas, and where they came from
Taken straight from the data file, unedited. Expressions are quoted verbatim — rewriting someone else's measurement to match a house style is how you break it.
| ID | Expression | Status | Source |
|---|---|---|---|
| F04 D 강화류 hit-type | D = 1 + multiMult + smashMult + aoeMult + comboMult + ultMult | MEASURED confidence HIGH | enban#134 S2 |
| F24 BUCKET RULE | Within a bucket: ADDITIVE. Between buckets: MULTIPLICATIVE. Damage-judgement types (crit / unguarded / extra-hit) multiply with each other. | MEASURED confidence HIGH | DC mabinogimobile#413525 S0-S2 |
| F26 연타+강타 SHARED POOL | effective = (smashStat + multiStat) / 8500 — the two stats are interchangeable, only their SUM matters | MEASURED confidence HIGH | DC mabinogimobile#821060 2025-05-13 S1 |
| F05 D.연타 multi | multiMult = (1 + multiStat/8500) * (1 + multiDmgPct) - 1 | MEASURED confidence HIGH | enban#134;mobis2stat S2 |
| F06 D.강타 smash | smashMult = (1 + smashStat/8500) * (1 + smashDmgPct) - 1 | MEASURED confidence HIGH | enban#134;mobis2stat S2 |
| F07 D.광역 aoe | aoeMult = (1 + aoeStat/8500) * (1 + aoeDmgPct) - 1 | MEASURED confidence MED-HIGH | enban#134 S2 |
| F08 D.콤보 combo | comboMult = ((1 + comboStat/17500) * (1 + comboDmgPct) - 1) * (n/4); n = 1 at 1+ combo, 2 at 10+, 3 at 50+, 4 at 100+ | MEASURED confidence MED-HIGH | enban#134 S2 |
| F09 D.궁극기 ult | ultMult = ultStat/8750 | MEASURED confidence MED | enban#134 S2 |
| F03 C 피해증가 dmg-inc | C = ((1 + swi/8500) * (1 + skillDmgPct) + helioPct + itemDmgPct + crescendoPct + artiDmgPct + synergyDmgPct) * (1 + armorBreakPct + synergyTakenDmgPct) | MEASURED confidence HIGH (math) / MED (label) | enban#134;mobis2stat;enban#599072 S2 |
| F25 NOT MODELLED | 크로스(cross), 약점(weakness), 분노상태(enraged), 카운터상태(counter-state) are explicitly unmeasured | MEASURED confidence HIGH | enban#134 S2 |
MEASURED means somebody hit something and recorded the number. DERIVED means it was worked out from other rows. The confidence values are the original compiler's own, not ours.
What this cannot tell you
- Cross, weakness, enraged state and counter state are explicitly unmeasured by the original compiler, so nothing is invented for them here.
- Per-skill base coefficients have never been published. That is why a real damage number for one specific skill is not obtainable.
- Taiwanese values may differ from Korea's. This page describes the current Korean build, and there is no official Taiwanese formula to check it against.
Data file generated 2026-08-18. Nexon publishes stat values but never formulas. Anything without an official announcement behind it is a player estimate, and this site does not label it official.
Cast-speed glossary: skill speed, casting/charge speed and skill-use speed are three different things
Taiwan board posts routinely compare these three as if they were one "speed". They act on different parts of a skill. The game splits every skill into a preparation phase (casting, charging) and an execution phase (the attack motion, channelling). Skill speed accelerates both; casting/charge speed only the preparation; skill-use speed only the execution. None of the three shortens cooldowns.
Confident The two-phase definition is the official description (the TW tester's screenshot; the Korean gallery write-up quotes the same wording from the official site). It is not a player guess.
| Term (client wording) | Which phase it speeds up | Where it appears | Measurement and source |
|---|---|---|---|
| Skill speed 빠른 스킬 A character-sheet stat. TW client: 快速技能 — the same wording the official reroll-odds table uses. | Both phases — including instant skills with no casting/charge text. | Gear lines, the reroll pool, potions and other stat sources. It is a number, not a rune percentage. | From 372 to 1,400 (+1,028), the full animation of five skills shortened by 5.5%–12.1%. The tester summarised it as roughly 8–10% per +1,000, smaller on short skills, and apparently diminishing. Moderate |
| Casting / charge speed 캐스팅 및 차지 속도 Rune wording. TW client: 施法、蓄力速度 (this site's rune library renders it 詠唱及蓄力速度 — same line). | Preparation only: the casting or charging time while the skill icon and progress bar spin overhead. Not the attack motion after a charge completes, not channelling, not instant skills. | Supernatural (20%) / Supernatural+ (25%): 55% chance on a skill or basic attack, 5 s. The tester also lists Greatsword Warrior passive 1 (Onslaught). Same wording in the rune library: Mystery, Oddity, Frostwind+, Persistence+, Onrush+, Twin Stars; Consul's flaw is −8% the other way. | Tester's 18 Aug follow-up video: Supernatural shortens the casting/charge phase, Storm does not. Mage and Priest (Sanctuary, Holy Spear) tests in the comments agree. Bard skill 1 is not sped up by Supernatural — no casting text, just a long instant animation. No percentage was measured. Moderate |
| Skill-use speed (skill activation speed) 스킬 사용 속도 Rune wording. TW client: 技能施放速度 / 技能發動速度 (rune library: 技能使用速度 — same line). The "attack speed" on the same line means basic attacks. | Execution only: start-up, recovery and channelling tick rate. A channelled skill ticks faster but lands the same total hits (1 hit/s for 5 s = 5 hits; at +100% speed, 5 hits in 2.5 s). | Storm (32%) / Storm+ (40%): on an unguarded hit, 5 s. Same wording: Torrent (24% / 3 s), Sandstorm (10% always + 24%), Onrush+, Twin Stars; Greatsword Warrior passive 1 (Onslaught); Consul's flaw is −8%. | Tester's 18 Aug follow-up video: the Fighter's charged punch finishes charging at the same moment with or without Storm, but the Storm side's recovery ends first. No percentage was measured. Moderate |
The tester's table (372 vs 1,400 skill speed, 120 fps frame count)
| Skill | at 372 | at 1,400 | Difference | Shortened by |
|---|---|---|---|---|
| Rogue skill 1 | 0.608 s | 0.575 s | 0.033 s | 5.5% |
| Rogue skill 3 | 1.075 s | 0.967 s | 0.108 s | 10.1% |
| Musician skill 1 | 1.725 s | 1.583 s | 0.142 s | 8.2% |
| Musician skill 4 | 2.542 s | 2.233 s | 0.308 s | 12.1% |
| Musician skill 3（3rd hit） | 3.967 s | 3.617 s | 0.350 s | 8.8% |
Timed from the overhead skill icon appearing to disappearing. Channelled skills were excluded because the training dummy breaks too easily and the slow-motion effect distorts the timing. The tester wrote that no single rule fit the five numbers.
How to tell which speed a skill takes
- Skill text says "casting time" or "charge time" → it has a preparation phase, which takes casting/charge speed.
- The attack motion itself, and skills whose text says "channelling time" → execution phase, takes skill-use speed.
- None of the three phrases → instant. The whole skill is execution and takes only skill-use speed and skill speed. A long animation does not mean a preparation phase.
- All three only make the motion faster. Cooldowns are unaffected.
What the board picks (player opinion, not measurement)
- Supernatural is only worth it for classes heavy in casting/charging: mage classes have casting time on almost every skill, and Priest (Sanctuary, Holy Spear), Musician casting, the Fighter's charged punch and the Longbowman's two charged accessory skills were all reported as noticeable.
- Crossbowman has no casting/charge text in its kit, so the tester says no. Greatsword Warrior is split: some think an armour engraving slot is too much to spend on it, others run it happily with Momentum.
- Storm works for every class but needs an unguarded hit to trigger, so uptime is poor on big bosses that rarely break — the most consistent reservation in the comments.
Unconfirmed Unconfirmed
- "Roughly 8–10% per +1,000" rests on two stat values (372, 1,400), five skills, and a 5.5%–12.1% spread. Whether the falloff is a curve or "time ÷ (1 + stat/10000)", as one comment relayed, nobody has measured.
- The tester, citing namu wiki, says speed percentages add with each other and with the percentage skill speed converts to, so a 32% Storm is worth about 3,000–5,000 skill speed. This site could not check that page, so the conversion stays a player relay.
- Greatsword Warrior passive 1 (Onslaught): the tester says it lists both casting/charge speed and skill activation speed; this site's Korean skill data has only skill-use speed. TW client text still to be compared.
- Whether the Bard's speed-up comes from Storm or Supernatural — the reporter wears both and could not tell.
- A Korean gallery write-up the tester cites calls "1% per 100" baseless; its author at 4,200 skill speed is not 42% faster. Same direction as the TW test, but no number.
- The TW client and board say 風暴; this site's rune library currently shows it as 暴風 (폭풍). Same rune.
Source: Bahamut Mabinogi Mobile board, 【密技】【瑪奇M研究所】快速技能與蓄力/施法速度 (2026-08-18, including the 18 Aug follow-up floor) and its comments. The Korean DC gallery post and namu entry the tester cites are marked as Korean-side relays. Open the thread →