127: Floor Six

By late May we had eighty-three members and had cleared through Floor 6 of the Black Castle.

The Floors 5 and 6 clears had each set new server records — Wenqing tracked them without fanfare, updated the guild's internal progress log, and the continental committee forum updated the records automatically. The Tianlong server forum added more threads to the accumulating list of threads about Severing Light. The forum's theory-crafting community had been trying to reverse-engineer our routing for three weeks; they were getting closer, which meant Floor 7 was where the server average would finally catch up to us, which meant Floor 7 was where the advantage we'd built from my knowledge of the layout would normalize. After that, the advantage would have to come from something other than foreknowledge.

By the third week of May the guild was ranked eighteenth on the Tianlong server's official guild rankings. The first probationary guild to break the top twenty. Wenqing had included that statistic in his weekly report without comment, which was his version of noting that it was significant. He didn't add analysis of what it meant for the guild's public position; the statistic was sufficient. If I wanted to think about the implications I could do it without his prompt.

The implications were: Severing Light was now visible in a way it hadn't been a month ago. Visible to the server's player community, visible to the continental committee's ranking observers, visible to Wang Jian's intelligence apparatus, and visible to the kind of players who were looking for a guild that was growing fast and clearing content ahead of the server average. The last category was useful. The others required management.

Wang Jian's response to the ranking was to send a second message to my inbox, this one through the official guild-message channel rather than the anonymous W-initial account he'd used before.

The message said: *Bladeless. Severing Light's ranking position is unusual for a probationary guild. I'd like to understand your recruitment strategy. Would you be open to a brief guild-cooperation discussion at your convenience.*

Guild-cooperation discussion. The phrasing was careful — not merger, not alliance, not absorption. Discussion. The kind of framing that was designed to seem neutral while the actual intent took shape behind it. Wang Jian was good at this kind of framing. He'd been doing it for the entire server's existence.

I read it and filed it under *later*. Later was a real time horizon — not dismissal, not avoidance. I would talk to Wang Jian again when talking to Wang Jian was the right move, and right now the right move was Floor 7.

***

The Floor 6 boss, the Ashen Empress, was a Lv 42 Magic-type with a formation-disruption ability that the server's advance teams had been struggling with for two weeks. The ability — **Ashen Cascade** — applied a randomized reposition to up to six party members within twenty meters every forty-five seconds. The Cascade didn't deal damage directly; it scattered formation arrangements without warning, moving players from their positions to random points within the zone's twenty-meter radius. The damage came from the scattered players losing their formation roles — healers out of range, tanks off-position, DPS in each other's line of fire.

The standard guilds had been solving Ashen Cascade by increasing their formation's redundancy — keeping duplicate coverage arrangements so that a random scatter of six didn't leave gaps that couldn't be quickly filled. This worked. It just wasn't efficient: redundant coverage reduced DPS output by approximately fifteen percent, which lengthened the kill window significantly.

TwilightTide's solution was different.

She'd brought it to the council session three days before the first attempt. Two pages of handwritten notes, the same format as her Ember Vault analysis — clean, dense, the specific density of a document that had been through at least two drafts before the final version was printed. The handwriting was small and legible throughout, not rushed.

The first page showed a heat-map of the Ashen Cascade's randomized repositioning. She'd built it from the server forum's published logs, aggregating the reposition data from the six teams that had attempted the Empress to date. "The algorithm isn't truly random," she'd written. "It's constrained random with a preference for mid-range distances. Players are repositioned eight to twelve meters, not one to twenty. The bias toward mid-range is consistent across all documented attempts — twelve percent variance at the extremes, eighty-eight percent of repositions landing in the eight-to-twelve-meter band."

The implication was in the second paragraph: "If your formation is already at eight-to-twelve-meter spacing, the repositioning has minimal disruption effect. Scattered players land in positions equivalent to where they started relative to the formation center."

The second page showed the formation adjustment. Spread to a uniform eleven-meter spacing at the start of each forty-five-second cycle. Hold until Ashen Cascade fired. Tighten back to the seven-meter combat spacing immediately after the scatter. The Cascade would fire and find that it had, essentially, rearranged a formation that was already arranged for rearrangement.

Wenqing read the analysis and spent four minutes verifying her math on the algorithm bias — I watched him cross-reference it against the forum's raw log data. "The bias estimate has a twelve percent variance," he said. "At the extremes of the range, three out of twenty players could be repositioned outside the disruption-neutral zone."

"Which three," Old Wolf said.

"Random selection from the reposition pool."

"Then we need a healer who can cover an outlier scatter," Old Wolf said. He looked at TwilightTide.

She was already looking back at him with the expression she used when she'd expected this question and had the answer ready. "I'll hold the outlier correction. Zhu Yuhan holds the standard formation. I run mobile coverage."

It worked on the second attempt. The first attempt failed because one outlier scatter landed two DPS members outside the disruption-neutral zone simultaneously, and TwilightTide couldn't cover both in time — the reaction gap was 0.3 seconds too wide. She logged the failure without frustration, filed the time-gap analysis, and told the council the correction before anyone asked for it: "I'll pre-position three meters closer to the statistical cluster of outlier positions. That closes the reaction gap."

"How do you know where the outlier cluster is," Old Wolf asked.

"From the first attempt. The Cascade's RNG seed isn't fully random — it's pseudo-random with a positional bias toward the south and east quadrants. I had one data point from tonight." A pause. "It's not a large sample. But it's a directional signal."

One data point. She'd made a positional adjustment from one data point because one data point was what she had and the adjustment was directionally correct.

The second attempt was clean. She held the outlier correction every cycle. The formation spread-tighten rhythm ran at the forty-five-second interval like a heartbeat. The DPS output stayed high because we weren't running redundant coverage. The kill window shortened.

[System: ASHEN EMPRESS eliminated. BLACK CASTLE FLOOR 6 CLEARED. Server record: 3 hours, 44 minutes. SEVERING LIGHT.]

The guild's highest Floor 6 record before our run had been 5 hours 11 minutes, set by a sixty-eight-member guild two weeks prior. We'd beaten it by 87 minutes.

Old Wolf watched the kill notification and said, in the quiet of the Central Hall's post-kill ambiance: "The Empress analysis."

TwilightTide: "Yes."

"You ran the reposition algorithm analysis on your own."

"Yes."

"How long did it take."

"About two hours." She said it plainly, as if two hours of algorithm analysis was an ordinary way to spend an evening. "I can run that kind of analysis quickly when I'm not sleeping."

"How often do you not sleep."

She thought about it. Not defensive — genuinely counting. "Four to five nights a week."

Old Wolf looked at me. I said nothing. He looked back at TwilightTide. "Can you sleep when you're not on tour."

The word just existed in the room for a moment. It had weight, but not the weight of exposure — more the weight of a thing named correctly after a period of being named approximately.

TwilightTide looked at Old Wolf. Old Wolf looked back at her with the expression he used when he'd concluded something and wasn't going to pretend he hadn't.

"My management team says I'm too busy," she said, eventually. There was something careful in the phrasing — not evasion, but precision. "I'm actually too quiet. The tour is — loud enough that I can sleep. The gaps between are too quiet. I have too much room in my own head."

"So you run protocols."

"So I run protocols."

"And in the quiet you produce analysis that a full-time strategist would take a week to build."

"The alternative is lying awake." She said it without self-pity. It was just the arithmetic of the situation: she couldn't sleep in the quiet, so she worked. The work was good. The quiet was still too quiet. Those things could both be true simultaneously.

Old Wolf nodded slowly. He looked at the Floor 6 kill notification, at the server record that had just changed to a number with Severing Light's name next to it. "The Empress record is ours by ninety minutes," he said. "Because of the analysis." He picked up his empty tankard from the stone bench where he always left it during combat. "Get some sleep when you can."

He walked to the outer gate.

Wanqing, from her position at the east formation line, looked at TwilightTide.

TwilightTide looked back at her.

"He knows," Wanqing said.

"I know he knows."

Wanqing was quiet for a moment — the kind of quiet she used when she was deciding how to say something. "The quiet part isn't about the tour," she said, gently. "Is it."

A long pause. The Central Hall's post-kill ambiance was the particular stillness the game engine generated after a boss clear — the environmental audio dropped to the floor's lowest register, the combat lighting gone, the torch-texture illumination soft and steady.

"No," TwilightTide said. "It isn't."

She didn't say what the quiet part was about. She didn't have to, and Wanqing didn't press her. She left the floor instance before anyone else, walking east toward the mountain gate at her standard long stride — head slightly down, the particular angle she moved at when she was done with a thing and ready to be somewhere else.

Zhu Yuhan said, from the rear formation position where she'd been standing quietly for the last five minutes: "She'll be at three AM protocols in four hours."

"Yes," I said.

"Someone should be there with her."

I looked at Zhu Yuhan. She was looking east, toward where TwilightTide had gone.

"I know," she said, before I could respond. "I'm usually there. I'll be there."

She walked east after TwilightTide.

I stayed by the Floor 6 kill marker for a few minutes. The Black Castle's ambient dungeon sound was the only thing in the Central Hall of Floor 6 — the low constant hum of the game engine's environmental audio running at its minimum, the atmospheric register that spaces had when they'd been emptied of everything that had been in them.

Eighty-three guild members. The two who ran protocols at three AM because the quiet was too quiet were heading east together. The server-first record board had Severing Light at the top of four consecutive floors. Wang Jian's recruitment-strategy inquiry was still in my inbox, filed under *later*.

I thought about TwilightTide's answer to Wanqing: *No. It isn't.* The quiet part wasn't about the tour. She hadn't said what it was about. Wanqing hadn't asked. There was a specific quality to that exchange — both of them understanding that the naming of it wasn't the point, that the acknowledgment was the thing, and the acknowledgment had been made and that was sufficient.

That was one of the things Wanqing did well. She created space for things to be partially said without pressing for the full sentence. It was a skill that required more restraint than it looked like. Most people, when they sensed something was close to the surface, pushed. Wanqing waited.

I didn't know what the quiet part was about. I had a hypothesis, the same way I had a hypothesis about most things I couldn't confirm yet. I'd hold the hypothesis and continue.

The Floor 6 kill notification was still active in my interface. Wenqing would file the clear time in the server records tonight and the forum would have threads in the morning. I'd check the threads in the morning and see what they said about TwilightTide's positioning, which would give me a read on how much the server was beginning to understand about her analysis methods.

I closed the guild channel and left the floor.