103: Formation Practice

Wenqing had three alternate formations ready by Thursday. He presented them Friday evening at the south gate practice grounds — a stripped-down instance map he'd requisitioned through the guild's continental committee access flag, sized for twelve players and tiled in flat grey stone so the combat geometry wasn't contaminated by terrain variables.

I walked into it at seven PM and found the three alternate formations already drawn in chalk on the ground. The flat grey stone caught the evening's ambient light without shadows, which was the point — practice geometry had to be readable from every angle and at every level of attention. Wenqing had drawn the formations with the precise hand he used for everything he considered important: clean lines, consistent scale, the directional markers aligned to true north within the instance's internal compass. The chalk was white against grey. The three configurations looked, at distance, like three different arguments for how twelve people should stand.

"Left chalk is the conventional anchor-plus-ranged formation," Wenqing said. He was standing to one side with his Summoner staff planted like a walking stick, an exact replica of the way he stood at the back of the HZUT library's reading room when he was explaining something he considered obvious. "Center chalk is the rotating-vanguard variant. Right chalk is what I'm calling the Scattered Fan."

Old Wolf stood over the right chalk. He studied it for a long moment. "This looks like a way to get three people killed simultaneously."

"It looks like that," Wenqing said. "It isn't. The Scattered Fan opens with a wide three-point spread that forces opposing DPS to commit to one of three attack corridors. Once they commit, two of the three points collapse inward, creating a pressure envelope that the anchor tank doesn't have to hold alone. The third point maintains the range threat."

"And if their DPS doesn't commit."

"Then they're standing still, which is worse for them than for us."

Old Wolf grunted. "Show me the collapse timing."

We spent two hours running the collapse timing with the guild roster we had — twelve members on a Saturday morning rotation. Mother Plum's Iron Plum anchored as second tank, slower than Old Wolf but solid on holding a corridor under pressure. Wanqing worked the east point of the Scattered Fan formation with precise, unhurried efficiency. Zhu Yuhan ran the healer position from behind the center fold, positioning herself at the six-meter mark Wenqing had specified.

The first two runs had the specific quality of new things — everyone understanding the plan conceptually, nobody yet understanding it with their bodies. Iron-of-Five overcommitted on the first approach and walked into his own team's collapse window. GreenSun held his arc angle correctly on the second run but went a full second late on the signal, which meant the pressure envelope formed with a gap on the right that the notional opposing DPS would have exploited immediately. Wenqing noted both without judgment. Old Wolf noted them with exactly the amount of judgment he always applied: enough to clarify what had gone wrong, not enough to make the person doing the analysis feel worse than useful.

It worked on the third run.

Not cleanly — there was a lag in the eastern collapse where Wanqing's pull-back timing was off by one second because she was accounting for the terrain slope that wasn't present on the flat grey stone — but the core mechanic held. The pressure envelope formed. Old Wolf said "that's a real thing" in the voice he used for things he was willing to die with.

The fourth and fifth runs identified two more variables: the center-fold timing when the convergence started from unequal distances (runs four and five both broke on this), and the rear-line positioning when the DPS cluster moved earlier than projected. On run five, Iron-of-Five overshot his diagonal by three meters and tangled with GreenSun's approach line. They stopped. Fixed it. Ran it again. Small adjustments built like sediment.

After the sixth run, we sat at the edge of the practice instance to eat. Iron Plum had brought a slab of dried pork from the guild crafting stores. Mother Plum, who had been watching the drills from the north wall with her paper fan closed and tucked in her belt, said: "The sixth run was clean."

"The sixth run had the slope correction built in," Wanqing said.

"Yes. Which is why the sixth run was clean and the first five were information."

Wanqing looked at her. "You've done formation drills before."

"My cousin and I ran the civilian guild's NPC-relations training cycles for two years." Mother Plum opened her fan, considered something on its painted surface, closed it again. "Formation drills and NPC-relations work have the same requirement: you run it until the correction is in your feet, not in your head. Then you run it once more to be sure."

Old Wolf, eating dried pork with the deliberate calm of a man who only ate when the thinking was done: "She's right. Once more."

We ran it a seventh time.

***

*Ding!* [System: Level Up! Lv 30 → Lv 31] [Berserker Class Bonus: STR +4, VIT +3. New skill available: Rending Fury — Lv 1] [Rending Fury — Lv 1: A two-hit forward strike that deals 180% physical damage on first hit and 220% on the second. Applies 3-second Rend debuff (DEF -15%). Cooldown: 12s. Mastery: 0/100.]

It happened on the eighth pack on Saturday morning at five-forty AM — Zhu Yuhan was at her usual session and I'd logged in early to catch the low-concurrency window. The level notification rolled over on a pack of Lv 30 bandit lieutenants at the ridge's eastern approach, and I felt the stat adjustment the way you feel a coat settling onto your shoulders.

I opened my stat sheet.

[Character: Bladeless] Class: Berserker (Lv 31) HP: 3,240/3,240 | MP: 410/410 ATK 312 | DEF 185 | STR 224 | AGI 143 | INT 38 | END 196 Equipped: Iron-Tide Battleaxe (Blue, ATK +92 STR +18), Stonewall Greaves (Blue, DEF +24 END +12), Iron-Crown Gorget (Blue, DEF +18 STR +8) Skills: Crescent Moon Slash (Lv 10+, hidden mastery), Rending Fury (Lv 1), Ironbody Stance (Lv 7), Severing Form (Lv 3) Gold: 340 silver, 18 copper

Four levels to go.

"Congratulations," Zhu Yuhan said. She was checking her healing rotation log from the pack cycle. "You're ahead of the twenty-one day schedule by four hours."

"The Sunday runs will be longer."

"I've blocked six hours Sunday. Four hours Monday through Friday, six hours Saturday and Sunday."

I looked at her. "That's thirty-four hours a week."

"The medical coursework is lighter in February. I'll adjust when it picks up."

"What is your father's specialty."

"Cardiothoracic surgery." A pause. "He's on call most weekends. My mother works Saturdays. The house is quiet. I play."

I didn't push further. In the old timeline I'd known several players who logged in from quiet houses during odd hours, and the reason was usually the same: they'd built an inside life because the outside one had a particular texture they couldn't change. Zhu Yuhan was efficient and precise and had learned to build what she needed with available materials. The summons scroll. The solo protocols. The annotated kill-cam printout she'd brought to the recruitment meeting.

She didn't need my analysis of it.

"I'll be here Sunday at three," I said.

"I'll be here at three," she said.

***

Saturday afternoon I ran the alternate formation drill data to Wenqing for refinement. His reply came back in forty minutes, annotated in red: three adjustments to the collapse timing, a note about the slope-correction variable needing a conditional trigger based on terrain type, and a single line at the bottom that read: *Azure Tide's Round 2 match will occur on a Flatland League terrain map. No slope variables. Scattered Fan as drilled will perform at 94 percent efficiency.*

I forwarded it to Old Wolf.

Old Wolf's reply: *94 percent. Run it again Sunday to be sure.*

I sat with the 94 percent figure for a while before replying. In my experience with Wenqing's analysis framework — which was careful, methodical, and tended to be conservative on estimates — a 94 percent efficiency projection on flat terrain was as good as a projection got before the match made it real or not real. He didn't round. He didn't soften. When he said 94, he meant the number was defensible from the data and the 6 percent gap was specific rather than vague. That was the part worth examining: what lived in the 6 percent.

Ninety-four percent was a number that felt comfortable and therefore required watching. In my experience — five years of old-timeline and three months of new-timeline — the formations that performed at 94 percent in analysis performed somewhere between 85 and 97 percent in practice, depending on which 6 percent they'd missed. The gap wasn't random variance. It was always a specific thing that hadn't been in the model — a terrain detail, a timing interaction between two players whose individual patterns were known but whose combined pattern wasn't, a decision point where the plan assumed one human behavior and got a different one. The map was not the territory. That was not a criticism of Wenqing's model, which was the best available. It was a description of the fundamental limitation of modeling human behavior at high precision. The question was always what the 6 percent was. Wenqing's terrain-type analysis was sound. What it couldn't account for was the human variable: the moment someone on the east flank moved half a second early because they heard something they weren't supposed to hear, or the moment the opposing DPS didn't follow the predicted path because their guild leader made a call in the last thirty seconds before the match that no analysis could have anticipated. I'd seen formations fail on smaller margins than 6 percent.

I filed the uncertainty and noted to run it again Sunday as requested.

Saturday evening, Wanqing came to the Greenleaf Inn with a packet of spiced sunflower seeds and sat across from me and said: "Wenqing ran the bracket projection again."

"And."

"Golden Serpent won their Round 2 match this morning." She cracked a seed. "Against a Lv 31-average guild. They ran the exact counter-formation to the Scattered Fan."

I held that for a moment. "They didn't have our drill data."

"No. But they had the standard anchor-plus-ranged formation data from Round 1, and they extrapolated three most-likely variations. The Scattered Fan is one of those three variations. They tested all three." She paused. "Xu Ming — Golden Serpent's leader — is not a casual guild commander. He's run counter-formation testing against unconfirmed opposition variations. That's not bracket strategy. That's intelligence work."

"Wang Jian didn't promote him from the vendor contract because he was a good vendor," I said.

"No." She cracked another seed. "He promoted him because he knows how to think about incomplete information."

We sat with that for a moment.

"Round 2 in eight days," she said. "We have three formations. Azure Tide expects the standard anchor. We run Scattered Fan from the first position."

"Yes."

"And if Golden Serpent's Round 2 analysis reaches Wang Jian before our Round 2 match."

"It reaches him Friday night at the earliest. Our match is Saturday at six AM. He'll have eighteen hours to adjust Xu Ming's Round 3 preparation. Not enough time for a full counter-development."

She considered this. "You're counting on Wang Jian moving at a human pace."

"I'm counting on an eighteen-hour lag," I said. "That's not Wang Jian's pace. That's the pace of his communication chain."

She looked at me. "That's a specific piece of knowledge."

"Yes," I said. "It is."

She cracked a third seed. She looked at the shell for a moment. In the lamplight from the window her expression had the quality it had when she was filing rather than concluding — taking in information and choosing to hold it in an unfiled state rather than forcing a shape onto it prematurely. She was good at holding things in an unfiled state. I'd come to think of it as one of her core professional skills: the patience to sit with an incomplete picture without papering over the incompleteness with a wrong conclusion. She said: "I don't ask."

"I know."

"I notice," she said. "I file. I don't ask."

"I know that too."

There was a particular quality to that silence. Outside the Greenleaf's window a cart vendor was arguing with a customer in the muffled way of an argument conducted under a noise ordinance — present but without detail, just two raised registers that eventually resolved into the particular sound of a sale made under mild protest.

In the old timeline, the version of Wanqing I'd known had been someone I'd never had the conversation with — the conversation that established what we both knew I was holding and what she'd decided to do about it. She'd died without us having that conversation. I'd spent five years regretting it the way you regret a door you left unopened. In this timeline we hadn't had the conversation either, but we'd arrived at the same understanding by a different path, which was the part that mattered.

She pushed the packet of sunflower seeds across the table toward me. "Eat. You've been running formation drills for two days and I can tell you skipped dinner."

I ate the sunflower seeds.

Outside the Greenleaf's narrow second-floor window the Jianghai late-February sky was the pale grey of the in-game rendering engine's winter cycle. Below the window, the street was quiet at this hour — the cart vendors who worked the evening rush had mostly packed up, leaving the cobblestones to the stragglers and the lamplight. The inn itself was full on the lower floor, the regular noise of a game-world tavern running at standard capacity. Up here in the private room it was just the two of us and the seeds and the eight days remaining.

Eight days from now we'd run Scattered Fan against Azure Tide, and Wang Jian would watch the kill-cam, and somewhere in the Tianxia Coalition's operation chain the name "Bladeless" would move one rank up the list he kept.

I cracked a seed.

Four levels to go.