140: The Bracket Draw

The CW I bracket draw was Sunday August 16 at nine AM server time.

I watched it in the east computer lab with Wenqing — not in-game, through the continental committee's public event broadcast, which ran on a dedicated web channel with the specific production quality of official events that were important but not quite important enough to warrant a full production team. The broadcast had one camera angle and two commentators reading from prepared notes. It was sufficient.

The draw was conducted in the committee's ceremonial format: a randomized seed assignment that placed sixty-four registered guilds into a double-elimination bracket, four rounds per side, with the top eight finishers from each group advancing to the championship bracket. The ceremony itself took twenty-two minutes, which was a long time to watch names appear in bracket positions, but I had Wenqing's probability model open in the adjacent window and the waiting had a purpose. The computer lab in the August morning was quiet — most of the student population was on summer schedules, sleeping in, or off campus entirely. The east lab had the particular morning quality of a space running for a handful of people who'd arranged to be somewhere specific at a specific hour for a specific reason. Wenqing was one of them. He had his simulation window open and was running pre-draw calculations without comment.

His model had taken him three days and was a twelve-variable simulation that accounted for each registered guild's floor-clear records, average roster performance metrics, known formation tendencies, and the structural advantages of coalition-affiliated guilds in multi-day bracket formats. He'd explained the variable weighting to me on Friday and I'd asked three questions about the formation-tendency coefficients, which he'd answered with the patient specificity he used when he was confident in the methodology.

His model's input for Severing Light: 100-member roster, full-charter, Floor 8 clear record (server best on five of six floors cleared), CW I registration as independent.

His model's output for Severing Light: **34% probability of top-five finish. 58% probability of top-ten.** The top-five probability was lower than I wanted and higher than the forum's 60/40 consensus suggested, because the forum's 60/40 included favorable assumptions about the bracket draw that the model didn't make without seeing the actual positions.

The draw took twenty-two minutes. Sixty-four guilds, randomized, bracket positions assigned in sequence.

Severing Light's position appeared on the broadcast at the twenty-minute mark: **Group B, Seed 3.**

Wenqing's hands were already on the keyboard. He updated his model with the draw results and ran the simulation. The output came back in forty seconds — faster than I expected, which meant he'd pre-computed conditional outputs for each possible bracket position and was assembling rather than calculating.

**Top-five probability: 61%. Top-three probability: 28%. Championship bracket probability: 89%.**

The Group B draw had put three of Tianxia Coalition's four sub-units in Group A. The fourth coalition sub-unit was in Group B with us, but it was Seed 8, which meant we'd face them in the second round at the earliest and the fourth round at the latest, depending on the path. The other Group B guilds included two floor-5-clear independents and two smaller coalition-affiliated guilds whose formation depth was below Severing Light's benchmarks.

I looked at the championship bracket path and did the arithmetic while Wenqing's model confirmed what I was already seeing.

"The Group A draw," I said.

"Yes. Three coalition sub-units in one group means they eliminate each other before the championship bracket. At least one — probably two — coalition sub-units don't advance past Group A."

"Wang Jian will see this."

"He's already seen it." Wenqing looked at the probability output with the expression of someone checking whether the numbers felt right, which was a separate process from whether they were calculated correctly. "His championship bracket path requires at least two of his coalition sub-units to advance, which means he's competing against himself in Group A."

I looked at the bracket structure on the screen. Wang Jian had built a coalition architecture that dominated the server for six months, and the randomized draw had placed three quarters of that architecture in direct competition with itself before the championship stage. The randomization was genuinely random — no one could have arranged this. It was the bracket draw being the bracket draw.

"He's going to reconfigure his coalition before the bracket starts," I said.

"Merge two sub-units into one, reduce the Group A field."

"It'll reduce his Group A seed count. But it strengthens the unit that survives the merge."

Wenqing ran a second simulation with that parameter loaded. "If he merges two of three Group A sub-units before the bracket, his surviving merged unit's championship probability increases by eighteen percent. Our top-five probability stays at 61%."

"He'll do it," I said. "He's patient enough to sacrifice the bracket architecture to strengthen the surviving unit. The bracket architecture is this season's problem. The merged unit's strength carries into next season."

Wang Jian always played the longer game inside the shorter game.

Wenqing updated the simulation notes and filed the analysis in the guild's planning records. "The merged unit's formation will be approximately 180 members in the bracket configuration. We're at 100."

"I know."

"The depth differential matters in the four-day bracket format."

"I know."

"Our advantage is formation precision over formation depth."

"Yes."

He looked at the simulation output. "The 61% is real," he said. Not as a reassurance — Wenqing didn't do reassurance as a category — but as a factual statement about the model's reliability. "It's not optimistic math. The formation precision advantage has been documented in eight consecutive server-record performances."

"I know."

"You're not concerned about the 39%."

"I'm aware of the 39%."

"That's not the same thing."

I looked at the bracket draw on the broadcast window. Sixty-four guild names in a randomized arrangement, our name in Group B Seed 3 position. The 39% was the version of the tournament where we didn't place top-five. It was a real probability. It existed.

"The 39% is the version where we lose in the championship bracket," I said. "Not the version where we don't reach the championship bracket. The championship probability is 89%."

"Yes."

"The 39% is for the championship stage. We'll see what the formation looks like when we get there."

He filed the simulation in the guild's planning records and sent the bracket draw summary to the council channel.

***

Wanqing called at ten AM.

She was in Suzhou for the weekend — she'd taken the Saturday train down for a family obligation, and I could hear the Pingjiang Road ambient sound in the background when she picked up. The particular morning-quality of the canal street, which I'd been listening to since childhood and could identify from the specific combination of water sound and stone acoustic and food stall movement at the distance her family's shop sat from the canal.

"The draw," she said.

"Group B, Seed 3. Three coalition units in Group A."

"He's going to merge."

"Yes. Two sub-units, probably. Creates one 180-member merged unit for the championship bracket path."

"How long do you have before the bracket starts."

"Three weeks. The bracket runs September 6 through 9."

"Four days."

"Yes. Double elimination, nine rounds total per side."

She was quiet for a moment. The canal street continued in the background — the specific Sunday-morning quality that came from the market opening and the first tourists arriving and the food stalls warming up their equipment. I'd been hearing that sound at the canal since I was small enough that the canal wall was taller than I was. It wasn't homesickness, exactly, to hear it now. It was something adjacent to homesickness that didn't have a precise name. The sound carried its own version of the fund calculation and the bracket draw and the things I'd been building for nine months — all of that had the canal street somewhere in it, the shop three floors down from the flat, the accounts book on the kitchen table. The bracket draw was in the east computer lab in Hangzhou. The reason for the bracket draw was on Pingjiang Road.

"Xiaoyu wants to tell you something," she said. "She's been ready for two weeks. She was waiting for the bracket draw to be public information."

I was quiet.

"She asked me to tell you: she's ready. Her plan is ready. She wants to talk to you on the Sunday before the bracket — September 5."

"The Sunday before."

"Yes. She was very specific about September 5. She said *before, not after. I don't want the bracket to be over before I tell him.*"

I thought about Xiaoyu doing her homework at the staircase landing with the red-string bracelet on her left wrist, the light on the third step that was better than the light in the second room. She'd been wearing the bracelet since February. She'd been wearing the same expression since February — watchful, patient, building something in the evenings that I hadn't been allowed to see yet.

"I'll be in Suzhou on September 5," I said.

"I know," Wanqing said. "I told her you would be."

The canal street sound continued. A food stall vendor saying something to a customer, the casual professional tone of someone who'd been having the same morning exchange for years.

"She's going to be all right," Wanqing said.

"I know."

"I mean she was always going to be all right. She decided in February that she was going to be all right and she built a plan and the plan is ready. I'm just telling you so you're ready too."

I sat with that.

"All right," I said.

"The bracket draw is good," she said. "61% on Wenqing's model."

"How do you know Wenqing's number."

"He sent the council summary to the guild channel. I'm in the guild. I get the council summaries." She said it with the dry patience of someone explaining an obvious thing that the other person had somehow overlooked.

She'd joined the Severing Light guild roster in April — the civic-affiliate status that gave her access to the non-combat support role on the guild's public record. She received the council summaries because she was formally in the guild's communication structure.

I'd forgotten that I'd added her to the roster. That was the kind of detail that slipped when you were managing three AM sessions and floor-clear preparation and financial documentation and charter reviews simultaneously. The detail had slipped. She'd been quietly in the guild channel for four months.

"September 5 for Xiaoyu," I said. "September 6 the bracket starts."

"Yes. She thought about that. She said: 'Before, not after. I don't want the bracket to be over before I tell him.'"

"I'll be in Suzhou on September 5."

"I know," Wanqing said. "I know you will."

The canal street sound continued until she ended the call.

I sat in the east computer lab and looked at the bracket draw broadcast window. Sixty-four guilds, Group B Seed 3, three weeks to the start. Wenqing had closed his laptop and was already annotating the opponent analysis files for Group B's Seed 14 — our Round 1 match.

The draw was good. The work was the same as it had always been.

I pulled up the Floor 9 formation notes and started reviewing TwilightTide's latest analysis annotations. Three weeks. Enough time to drill whatever the bracket required, if we used the three weeks correctly.

The Floor 9 analysis was sixteen pages, which was TwilightTide's longest analysis to date. The Ironveil Sovereign's positioning-memory mechanic had more variables than the previous floors — more movement states, more targeting conditions, more edge cases where the counter-solution could break down. She'd annotated the edge cases in red. There were seven of them, each with its own contingency note.

I read the contingency notes.

She'd been thinking about this floor for longer than the analysis document's creation date. Some of the contingency thinking was in the language of someone who'd been running scenarios for weeks, not days. The mental architecture of a problem she'd had in the background of other things. I'd learned to recognize that quality in her written analysis — the contingency notes that were too precisely worded to have been drafted the same day, that had the slightly different sentence rhythm of thinking that had been refined across multiple sittings.

The analysis asked things of every member that the previous floors had built toward in different ways. Sixteen pages of the specific kind of complexity that was only visible to someone who'd been watching the formation develop since March and could see which new requirement was a new problem and which was an existing capability applied to a new context.

The Floor 9 attempt was August 22. That was seventeen days. Enough to drill the continuous-rotation mechanic to the point where the rotation was automatic rather than deliberate — where people moved without thinking about whether they were moving, because the habit had been built.

That was the training requirement for Floor 9. Seventeen days to build the habit.

I sent the analysis to the formation-drill calendar and scheduled the first practice session for Thursday.