Silence Duration Decay: Resists Count, Time Heals

The Japanese wiki has long documented a Silence-specific behaviour: keep casting Silence on the same monster and the duration gets shorter. As far as I can find, this has never been documented in English - BG Wiki's Silence page lists a flat Duration: 120 seconds and nothing about any decay. I spent a week measuring it, and it turns out to be a precise, predictable mechanic.

From 15 recorded fights (logs read frame by frame, with Paralyze II cast in the same fight as a control):

Silence duration = base duration x 0.968 ^ (effective resist count)

- Every resist before the spell lands costs 3.2%, multiplicatively.
- The loss is not permanent. It recovers at roughly one resist per minute, with no floor - leave it alone and it goes back to full.
- Landing the spell again costs nothing. Wait for it to wear off, recast, and you get the full duration every time. The counter is tracked per spell.

The formula reproduces all 15 fights to an average error of about 0.3% (worst case 1.3%).

This also confirms what BG Wiki says about Immunobreak: the break itself does not affect duration. What costs duration is each resist. "It gets shorter the more you cast it" simply means you are still casting because it is still being resisted.

Full write-up, with the log data and unlisted footage of all 15 fights:

Part 1 (the mystery): https://jinko0914.seesaa.net/article/521249192.html
Part 2 (the answer): https://jinko0914.seesaa.net/article/521302994.html

Two readers of the Japanese edition independently posted their own numbers in the comments; both land on the x0.968 ladder within about a second.

Nine resists and above is territory nobody has measured yet. If you feel like testing any of this, I would love to see your numbers - "it did not happen for me" is worth just as much.

I don't believe this behavior is Silence-specific. When I do Gaol RP farming involving fighting the same Atonement 3 NM three times, I'll typically have a RDM in party for one of the fights to keep Gravity/Bind on as Helix takes off 5%. Gravity's duration will start shortening throughout the fight after repeated casts, particularly on Arebati since it tends to take a few attempts to get it to land. Presumably the resisted attempts still count towards lowering the spell duration since I experience a less significant impact on Gravity duration reduction due to it always landing 1/1 on Mboze.

Maybe im misunderstanding, but enfeebs in general have a decay/resistance build on NMs afaik, though there being a recovery mechanic is new to me, always thought it was a lasting resistance to a debuff/s. Still good to get some hard numbers

Sterk - that Gravity pattern is exactly what the resist-count model predicts: on Arebati it takes several casts to land (resists piling up), while on Mboze it lands 1/1 and keeps full duration. Thanks - good to hear the same shape from another spell.

Nariont - I assumed the loss was permanent too. The recovery only surfaced because one fight had a 46-second gap before the spell landed, and it refused to fit the formula until time was in it.

One caveat on "enfeebs in general", because I think this is the useful part: the decay does not look general. On Arebati, Paralyze II also had to be resist-hacked before it landed - and it still came in at its full duration, while Silence in the same fight came in short. So this looks like a property of specific spells - Silence, Gravity, Bind and Sleep are the usual suspects - rather than a blanket NM resistance. And whether the spells that do decay all share the same numbers (x0.968 per resist, about one resist recovered per minute - measured only for Silence) is still open. If anyone measures one of the others, I would love to see the numbers.

A warning for anyone tempted to test those two, though: Gravity and Bind roll a random base duration even on a clean landing (roughly 30-120s for Gravity, 0-60s for Bind), so the per-resist loss would be buried in the roll - much harder to isolate than on Silence. If someone has already worked through a verification there, I would really like to hear how you did it.