RED MAGE - ENFEEBLING DURATION
🔇
How Many Times Did You Cast That Silence?
The mystery of a duration shorter than the numbers say
Paralyze II lands exactly as predicted. Only Silence comes up short,
and not by any amount a resist can produce.
and not by any amount a resist can produce.
ANSWEREDThe hypothesis in this article has since been tested against 15 measured fights. Read the follow-up ▶
Originally published 2026/7/30 (JP)
⚠️ Before you start: the second half is a hypothesis
The explanation I reach later in this article is a hypothesis built from my own measurements. It is not confirmed behaviour. Please read the measured numbers in the first half and the reasoning in the second half as two different things.
While grinding Arebati V25 in Sheol Gaol, one thing never sat right with me.
Silence runs shorter than its theoretical duration.
And not in a clean, readable way like "it got halved". It comes up awkwardly short. Every time I tried to explain it with resists, the numbers refused to line up.
I went a long time without an answer. I think I finally have one that holds together, but it is still a hypothesis. What I want to put on the table in this article is two problems.
1
The unknown multiplier
Silence has a documented behaviour where the duration shrinks if you keep casting it on the same monster. But no source anywhere states how much it shrinks.
2
Does a resist hack count as a "hit"?
On Arebati, the very first cast that lands is already short. It has never gone in even once, and the duration already looks like the tail end of a long series. The hypothesis: the resisted casts are being counted too.
⚠️ The second one really is only a hypothesis
I have observed it in exactly one environment: mine. That is not enough to call it fact.
Is it reproducible? Does the same thing happen on your setup? The more player reports come in, the more solid this gets. If you are willing, I would love for you to measure it too.
👇 What the rest of this article covers
1. A one-minute gap2. The duration table3. A control: Mboze4. Resists cannot explain it5. A Silence-specific rule6. Short from cast one7. Hypothesis: hacks count?8. What to do in practice9. The unknown multiplier10. Measure it with me
🕐 1. Paralyze II and Silence fall off a full minute apart
My Arebati V25 write-up puts it like this.
| Spell | Duration | |
|---|---|---|
| Paralyze II | 7:37 | Lands in full |
| Silence | about 6:20 | Cut short somehow |
These two spells should have exactly the same duration. Both have a base duration of 120 seconds. Same Red Mage, same gear, cast within seconds of each other in the same fight.
And yet they drift about a minute apart and fall off at different times.
That gap hit me directly in practice. You cannot put Saboteur on both when you refresh them, so I had no choice but to build the fight plan around which one gets it.
📊 2. Building a baseline first - the duration table
Before chasing the mystery I built a reference table. The target is an NM. "4 pieces / 5 pieces" is how many pieces of the Lethargy set I was wearing (Red Mage's reforged Empyrean armor, which carries Enfeebling Magic effect duration). "Saboteur" means the cast had Saboteur on it.
🔸 2-1. Silence / Paralyze II (120s base)
| Gear setup | Dls. Torque +1 | Dls. Torque +2 |
|---|---|---|
| Saboteur, 4 pieces | 7:19 | 7:37 |
| Saboteur, 5 pieces | 8:08 | 8:28 |
| No Saboteur, 4 pieces | 5:44 | 5:59 |
| No Saboteur, 5 pieces | 6:22 | 6:38 |
💡 About the neck Dls. Torque is how the client truncates Duelist's Torque, the Red Mage JSE neck. The +1 augments to Enf. Mag. eff. dur. +20% and the +2 to +25% - that difference is the entire gap between the two columns above.
The "Paralyze II, 7:37" from my Arebati article lands exactly on "Dls. Torque +2, 4 pieces, with Saboteur".
So the table is right, and Paralyze II behaves exactly as predicted. Nothing is wrong so far.
🔸 2-2. Other enfeebles (Dls. Torque +2)
| Spell (base duration) | Saboteur, 4 pcs | Saboteur, 5 pcs |
|---|---|---|
| Break, Bind, Gravity (30s) | 3:13 | 3:35 |
| Sleep, Sleepga (60s) | 4:41 | 5:13 |
| Sleep II, Sleepga II (90s) | 6:09 | 6:50 |
| Paralyze II, Silence (120s) | 7:37 | 8:28 |
| Slow II, Addle II (180s) | 10:33 | 11:44 |
| Distract III, Frazzle III (300s) | 16:25 | 18:15 |
| Spell (base duration) | No Saboteur, 4 pcs | No Saboteur, 5 pcs |
|---|---|---|
| Break, Bind, Gravity (30s) | 2:49 | 3:08 |
| Sleep, Sleepga (60s) | 3:52 | 4:18 |
| Sleep II, Sleepga II (90s) | 4:55 | 5:28 |
| Paralyze II, Silence (120s) | 5:59 | 6:38 |
| Slow II, Addle II (180s) | 8:05 | 8:59 |
| Distract III, Frazzle III (300s) | 12:18 | 13:40 |
💡 Note Bind rolls a base of 0-60s and Gravity / Gravity II roll 30-120s, so treat those two rows as a rough guide rather than a fixed value.
🧪 3. A control - on Mboze the numbers came out exactly right
I checked the table against my own footage: an Mboze run where I played Red Mage. Here, Silence landed on the very first cast - which means not a single resist hack went in beforehand.
796 left - Minelba casts Silence on Mboze. ← landed on the first cast
790 left - Minelba's Silence takes effect.
338.5 left - Mboze is no longer silenced. ← the moment it fell off
Translated from the Japanese client, so the wording is not verbatim English client text. "left" is the Sheol Gaol clock counting down.
| Seconds | ||
|---|---|---|
| Measured | 451.5 | 7m 31.5s |
| Theoretical | 457 | 7m 37s |
| Difference | 5.5 | = a match |
💡 Note A drift of a few seconds caused by when the log lines are emitted shows up in other players' published measurements too.
When Silence goes in on a single cast, it lasts exactly as long as the table says.
That became my baseline.
🤔 4. On Arebati it is short - and resists cannot explain it
Same Red Mage, same gear, same spell. On Arebati it comes out at about 6m 20s (380 seconds). Against a theoretical 457 seconds, that is a ratio of 0.83.
When a resist cuts a duration, it cuts it in steps: 1/2, 1/4 (there is no 1/4 resist on enfeebling magic - I am only listing it to show what "stepped" means).
| If it were... | Duration | In reality |
|---|---|---|
| A half resist | 228s (3m 48s) | Not what happened |
| Landing in full | 457s (7m 37s) | Not that either |
| Measured | about 380s (6m 20s) | No step matches |
⚠️ Half resists do happen - separately
As noted in the update log of my Arebati article (2025/6/9), if the opening Silence happens to land after zero or one resist hack, it very often comes back as a half resist.
So there are two layers: too few hacks = half resist; enough hacks = it lands, but at 0.83x. Neither of them is a resist step.
📖 5. There is a Silence-specific rule - documented only in Japanese
The Japanese-language FFXI wiki (wiki.ffo.jp) has this under its Silence entry.
There is a rarely-noticed behaviour: if you keep casting Silence on the same monster, the duration gets shorter. From around the 30th cast the duration approaches zero, to the point where the "it lands" and "it wears off" lines appear at the same time.
My translation of the FF11 Yougo Jiten (FF11用語辞典) entry for Silence.
⚠️ If you only read English sources, this will be new to you
BG Wiki's Silence page lists a flat Duration: 120 seconds and says nothing about any decay. I could not find this behaviour documented in English anywhere.
So please take this section as established, not as my claim: the Japanese wiki has described this for years. The hypothesis I build later starts from here - it does not start from scratch.
So it is known behaviour. "The more you cast it, the more it erodes" - which is exactly why the result is not a clean 1/2 or 1/4 but an awkward number. That fits the character of 0.83.
💡 NM-only? It is not stated outright. A separate entry on the same wiki, Enfeebling Magic, says "on some NMs, repeatedly landing the same enfeeble reduces its duration". In practice, assuming it applies to NMs in general is enough. I do not know what happens on normal monsters, but Silence duration only ever matters in NM fights, so it is not worth worrying about.
❗ 6. Except the first landed cast is already short
This is where it stops adding up.
If the rule is "it shrinks as you keep casting", then the first one should be full length, and the shrinking should start from the second or third.
❗ But on Arebati the first cast that lands is already about 6m 20s
It has never gone in even once, and the duration already looks like the tail end of a long series. This is the part I could not explain for a very long time.
💡 7. Hypothesis - resist hacks are being counted as casts
What finally occurred to me was how Silence has to be applied on Arebati. It does not go in cleanly. My article says as much.
Landing Silence requires two or more resist hacks plus Chironic Hose.
A resist hack means deliberately casting a spell you expect to be resisted: each resisted cast knocks the target's resistance down a step and makes the next one more likely to stick. (The underlying mechanic is called Immunobreak in English sources.) So you cast it over and over until it goes in.
By the time the first Silence lands, you have already thrown several Silences at the target.
What if those resist-hack casts are being counted as "casts" by the duration rule?
That single idea explains both observations with one mechanism.
| Mboze | Arebati | |
|---|---|---|
| How Silence goes in | Lands on the 1st cast | Needs 2+ resist hacks |
| Casts thrown before it landed | 1 | several |
| Measured duration | 451.5s (theory 457s) | about 380s (0.83x) |
Same Red Mage, same gear, same spell. The only thing that differs is how many casts it took to land.
And in the Arebati kill footage I analysed, the opening contained eight Silence casts (Paralyze II had two).
🔸 7-1. As far as I can tell, nobody has written this down
I went looking, in both Japanese and English.
- The behaviour itself - that the duration shrinks if you keep casting - is documented on the Japanese wiki.
- But I could not find a single source that addresses whether resisted (i.e. failed) casts are counted.
- And English sources do not document the decay behaviour at all.
💡 The objection I expect BG Wiki's Immunobreak page states that an Immunobreak only lowers the resistance rank and does not affect duration. I think that is a different question. That statement is about the Immunobreak mechanic; my hypothesis is about whether Silence's own duration counter is also counting the resisted casts. Two separate systems.
🛠️ 8. In practice - not "avoid the shrink" but "know how much it shrinks"
You might think "then just use fewer resist hacks". You cannot.
| Hack count | Result | |
|---|---|---|
| Too few (0-1) | Half resist, near-guaranteed | Duration halved |
| Done properly (2+) | It lands | but shorter than theory |
⚠️ "Cast less to protect the duration" is not an available option
On a target that requires resist hacking, a half-hearted hack is the worst outcome of the three. The shrink itself cannot be avoided.
Which means what actually matters in practice is this.
1
Hack properly and make sure it lands
2
Then find out exactly how much duration you lose
3
→ and you can predict when to re-apply Silence
The only reason I could build a Saboteur plan on Arebati - does it go to Silence or to Paralyze II - is that I had the number: Silence falls off at about 6m 20s.
If you cannot avoid it, the only option left is to measure it accurately and build it into the plan.
🔍 9. The biggest problem - how much it shrinks is completely unmapped
This may be the real point of the article. It is written that the duration shrinks, but there is no source that says how much. A multiplier? A flat subtraction? Is there a floor? None of it is known.
For anyone building a fight plan, that is the part that hurts most.
🔸 9-1. One thing can be said, though
On Arebati it settled at about 6m 20s every single time.
If the amount of shrink were rolled randomly, the value would scatter from run to run. It did not - it kept landing in the same place, and that means:
the shrink is probably not random. Same conditions, same result.
In other words, this is the kind of behaviour you can pin down by measuring it. That part is encouraging.
🔸 9-2. Three things still unknown
| Unknown | ||
|---|---|---|
| 1 | The amount of the shrink itself | Multiplier or flat? Is there a floor? |
| 2 | Whether resisted casts are counted | This flips the meaning entirely |
| 3 | What resets the counter | Per fight? Per target? |
Number 2 is the big one. If failed casts count, resist hacking is a pure loss. If only landed casts count, hacking is simply the cost of getting the spell in.
Note that on the resistance side it is said that a successful enfeeble resets the accumulated hack count - but nobody knows whether the duration side behaves the same way.
💡 It is also a problem for my own hypothesis "About 6m 20s every time" is itself a question mark against the hypothesis. If the hack count wobbles a little from fight to fight and the result still lands on the same value, then it is not a simple "count x fixed amount" - there may be a cap, or some other rule entirely.
🤝 10. Would you measure this with me?
Honestly, verifying this properly is a slog. You have to hold the conditions steady and repeat it many times, and doing that alone eats more hours than I have.
So if this interests you at all, I would love for you to measure it. If you play Red Mage and use Silence, this is data you can pick up alongside your normal fights.
🔸 10-1. Why reports matter
The second theme of this article - whether a resist hack counts as a hit - is only a hypothesis.
I have observed it in exactly one environment: mine. Whether it is something peculiar to my setup, or behaviour that everyone would see, I do not know yet.
Is it reproducible? There is no way to answer that other than collecting reports. The more matching reports come in, the more solid the hypothesis gets. And a report saying "it did not happen for me" is worth exactly as much.
The comments here are fine. Even one more data point moves this forward a lot.
📌 11. Summary
| Finding | ||
|---|---|---|
| 1 | Paralyze II behaves exactly as predicted (7m 37s) | Matches the table |
| 2 | So does Silence, if it goes in on one cast | Measured 451.5s on Mboze |
| 3 | On Arebati it drops to about 6m 20s | Resists cannot explain it |
| 4 | Silence has a shrink-on-repeat rule | Known (Japanese sources only) |
| 5 | But it is short from the first landed cast | The known rule does not cover this |
| 6 | Are resist hacks being counted? | ⭐ Hypothesis |
| 7 | The shrink does not look random | = it can be measured |
No amount of magic accuracy makes Silence itself last longer. But if the duration depends on how many casts it took to land, then stacking Enfeebling Magic Skill and magic accuracy stops being only about getting the spell to stick.
🔸 11-1. What is left, in the end
1
The unknown multiplier
Nobody knows how much the duration shrinks. But it does not look random, which means it should be measurable.
2
Does a resist hack count as a "hit"?
Still only a hypothesis. Whether it reproduces cannot be answered until reports come in.
There is a great deal here that I still do not know.
If you have data - "here is what I got" - please tell me. "It did not happen for me" is just as welcome. Both kinds of report move this forward equally.
にほんブログ村
