Thousand Swords: how the +24% actually works

✍️ Alice・愛麗沙

The stack ladder

+4% per stack, cap 6, is the in-game rune text. WHICH classes count as “sword-using” is named nowhere officially and community rosters disagree, so this site lists none.Runes

That table is computed from the rune’s own dataset row at build time, so it cannot drift away from what the game says. The full effect text as it appears in the client is browsable at the rune database. Thousand Swords is a legendary weapon rune and every class can socket it — the classId field on its row is literally all.

The misinformation is worth recording, because the shape repeats

The sequence is a textbook case of AI-generated guide damage, so it is worth writing down rather than quietly correcting.

In a thread with more than twenty thousand views, someone pasted a generative-AI explanation as an authoritative answer. It stated confidently that an archer socketing this rune would see the 24% “become 0%”, and it padded the claim with invented specifics about which Korean players ran which builds. None of it was real.

The correction arrived fast, in two stages. First a player restated the actual rule — +4% per sword-using class at level 40. Then somebody simply ran the experiment: socketed it on a non-sword class and gained +339 attack power. The condition demonstrably fires on characters who never touch a sword.

The most quoted line from that thread became a small motto on this site: the AI is bluffing you. And the reason it is worth quoting is not the wrong number — it is that the answer packaged “sounds reasonable” as “verified”, right down to fabricated sources. A useful test for any guide, this one included: does it tell you which parts it does not know? The next section is that test applied to itself.

Which six classes? We refuse to answer, and that is the responsible answer

A six-class roster circulates in the community — the warrior-and-greatsword family, broadly. It sounds reasonable. Here is why we will not print it.

The in-game rune description names no class at all. It says “sword-using classes” and gives the 6-stack cap, and that is the whole of the official text. Meanwhile the rosters Korean testers assembled independently contradict each other — some include the knight line, some include Frostmage (whose weapon is a staff-blade, which is exactly the kind of edge case that keeps the argument alive). Every “confirmed list” in circulation is an inference wearing the clothes of a fact.

So instead of publishing a roster, we rewrote the decision rule into a form that does not need one. That is the next section.

Is it worth levelling for? Only because you were already going to

Levelling six classes to 40 specifically for this rune is a real cost in spirit traces, and on its own it is not an efficient purchase. But there is a clean convergence.

The cumulative-class-level title track — the one that leads to the 675-level “Versatile” title, worth +275 attack and +275 defence — already requires you to take a dozen-odd classes to around level 35. Push the ones that plausibly carry a sword to 40 as you go, and Thousand Swords’ stacks are a free by-product of a route you were walking anyway.

That is the roster-free decision rule: you are breadth-levelling regardless, so order the sword-looking classes first. Guessing wrong on one class costs you a few hundred traces of sequencing, not a wasted investment.

The scale, from Korean player measurements: a single class from 1 to 31 costs roughly 150 traces; 31 to 35 costs about 450 more. Reaching level 35 is close to free. What is expensive is every level after that — and the same trace budget poured into a single class’s 60 → 65 segment runs past seventy thousand. Breadth and depth differ by two orders of magnitude, and level 40 sits right at the tail end of the cheap side. That is precisely why it works as a by-product.

Three details that change how you plan

It is character-bound, and the title track is not. Korean databases record Thousand Swords as counting this character’s class-level achievements. The title system counts your account total, so levels you push on an alt still feed your titles. They do not feed this rune. If you want to serve both systems from one trace budget, do the breadth-levelling on your main — titles read the account total, Thousand Swords reads your main’s own record, and one pile of traces pays twice.

The bonus does not appear in your combat power number. Players have tested this: socket Thousand Swords and your CP panel does not move, while your damage does. It has quietly become this site’s best teaching example for why combat power is not strength — a pick-up group screening by CP cannot see this rune at all.

You do not have to “maintain” anything. Class levels are permanent once earned (this is in the official class-change documentation), the stacks read your achievement record, and you never need to switch back or equip anything in particular. Reach it once and it is yours.

Shelf life: Korea has already moved past it

On the Korean server, Thousand Swords has dropped out of the mainstream weapon setups in Season 2 — the physical-class first pick there is now Rock Blade. That does not change anything on Taiwan’s server, which is still in S0 and where Thousand Swords is the default weapon rune on most class sheets on this site.

What it does tell you is how to budget. This rune is a correct answer with an expiry date, roughly a season and a half out. Take it to legendary, take the free stacks that come with your title route, and do not pour a season’s worth of fusion material into pushing it beyond that.

A claim you can overturn: that the roster question is genuinely unanswerable from public sources today. A complete exclusion experiment — one character, one class at a time to 40, attack power recorded at each step — would settle it. If someone runs it, this page publishes the roster the same day, with credit.

Where to go next: the full rune catalogue is at /en/runes/, the trace cost of breadth-levelling is calculable at /en/subclass/, and the title track this rune rides along with starts at the titles guide.