CHAPTER 21 : The Ghost in the Court
Nekoma Gymnasium — June 16th, Monday, 4:15 PM
The ghost appeared at 4:15 PM during a standard three-on-three scrimmage, and it was nothing like Arisu had imagined.
The dream interface the previous night had presented Future Branches as clean — ghostly prediction lines extending from a suspended volleyball, each one a potential trajectory, each one color-coded by probability. Elegant. Controlled. The kind of interface that suggested the ability would feel like reading a well-organized spreadsheet of the future.
The reality was closer to having someone shove a second television into his field of vision while he was trying to watch the first one.
Kuroo loaded a spike from position four. Standard cross-court approach, arm drawn back, the same hitting form Arisu had studied during blocking sessions that felt like a lifetime ago — that first evening in the gym when Kuroo had hit thirty-seven balls and Arisu had blocked four and learned that deception was the foundation of everything.
Arisu activated Future Branches.
[Zone Architect] Future Branches active. 1 branch generated. Prediction window: 2.5 seconds. MS: 40 → 32.]
The court split.
Not physically — the floor stayed solid, the net stayed vertical, the players stayed corporeal. But overlaid on the real court, a translucent second version appeared. A ghost-Kuroo, one shade paler than reality, completed the spike approach two and a half seconds ahead of the real one. The ghost-ball left ghost-Kuroo's hand on a cross-court trajectory, arcing toward position five, landing three feet inside the baseline.
Cross. Position five. Landing at... the marker says three feet in. Move NOW.
Arisu's body shifted before his conscious mind finished processing. Feet sliding left, platform setting, weight dropping — the receive position for a cross-court spike into zone five. The real Kuroo's arm came forward. The real ball left his hand.
Cross. Position five. Three feet inside the baseline.
The prediction was right.
Arisu's platform met the ball at the exact point the ghost-image had marked. The receive went up clean — cleaner than any dig he'd ever produced against Kuroo's spikes, because he'd been in position before the ball was hit, with his platform set and his weight balanced and his body arranged to redirect force instead of absorbing it.
"Nice receive!" Lev shouted from the opposite side.
Kenma said nothing. But his eyes tracked Arisu for a half-second longer than the play warranted.
The prediction was right. The receive was clean. The cost was eight MS. And the visual overlay — the ghost-court superimposed on reality — left me dizzy for approximately one second after the play resolved. Processing two versions of the same court simultaneously is like trying to read two books at the same time. The brain can do it. The brain doesn't like it.
Next rally. Yamamoto's approach from the right side.
Arisu activated again.
[Zone Architect] Future Branches active. 1 branch generated. MS: 32 → 24.]
The ghost-Yamamoto completed his approach. Ghost-ball: line shot, position one, fast and flat.
Line. Position one. Shift right—
Yamamoto hit cross.
The ghost was wrong. The prediction had shown line with sixty percent confidence, and the forty percent uncertainty materialized as a cross-court kill that Arisu was completely out of position for because he'd committed to the line prediction before the real ball left Yamamoto's hand.
The ball hit the floor behind him. Point scored against his side.
"Misaki, what was that?" Yaku's voice from across the net. "You moved AWAY from the ball."
"Bad read."
Bad read is an understatement. The prediction showed line and I committed. Full body commitment — feet, platform, weight, all positioned for a ball that was going to a completely different zone. The sixty percent accuracy means that four out of every ten predictions will be wrong, and when a prediction is wrong, the result isn't just a missed ball — it's a player who moved to the wrong position with conviction, which is worse than not moving at all.
The system doesn't tell me WHICH predictions are the right ones. Every ghost-image arrives with the same visual confidence. There's no flashing red warning that says "this one's the forty percent." They all look real until the real ball proves them wrong.
He shook his head. The dizziness from two rapid activations was building — not the headache yet, but the precursor, the sense that his visual processing was overloaded from rendering two overlapping realities in quick succession.
Third activation. Four rallies later. Arisu waited until a critical moment — set point for the opposing side, a play where the stakes justified the eight-MS cost.
Inuoka set up for a quick from the center. Ghost-image activated.
[Zone Architect] Future Branches active. 1 branch generated. MS: 24 → 16.]
Ghost-Inuoka spiked straight down, inside the three-meter line. The prediction was definitive — straight down, no angle, no cross.
Arisu called the block. "STRAIGHT! STRAIGHT DOWN!"
The block went up. Inuoka spiked straight down. The ball hit the block and deflected back onto the offensive side. Point saved.
Right again. Two out of three. But the cost — twenty-four MS in three activations, and the fourth will drop me to eight, and the headache is already pressing against my right temple.
Fourth activation. He needed to know the cost ceiling. Needed to feel where the wall was before he hit it in a match that mattered.
Kai spiked from position two. Ghost-image: deep cross, position six.
[Zone Architect] Future Branches active. 1 branch generated. MS: 16 → 8. Warning: MS approaching advisory threshold.]
The ghost-Kai hit cross to position six. Arisu moved to six. Kai hit line. The ghost was wrong.
Ball on the floor. Point lost.
And the headache arrived. Not the gradual build of sustained zone-rule use — a sharp, focused pressure behind his right eye that pulsed once and settled into a steady throb. Four activations in fifteen minutes. Thirty-two MS burned. Two correct, two wrong. The ghost images were still overlapping with his real vision — faint afterimages of trajectories that hadn't happened, lingering in his peripheral vision like the remnants of a camera flash.
Okay. Okay okay okay. Data collected. Future Branches at Level seven: one branch, sixty percent accuracy, eight MS per use. Functional assessment: a sniper rifle. Not a machine gun. One shot, maybe two per set at current MS levels. Used at the right moment — set point, match point, the critical rally where one read can swing momentum — it's powerful enough to justify the cost. Used indiscriminately, it drains the tank in four shots and leaves me dizzy with a headache and a forty-percent chance of committing to the wrong position.
The ability I've been chasing since the dream interface showed it as a preview is HERE. It works. And it's as expensive and unreliable as every other tool the system has given me.
He sat on the bench and rubbed his temples with both hands. The scrimmage continued without him — Nekomata had noticed the quality fluctuation and waved him off the court with the particular half-lidded expression that meant "you're done for today, don't argue."
Kenma sat down next to him. Not immediately — five minutes later, after the scrimmage rotation ended and the team moved to serving drills. He held a water bottle and his phone, the standard Kenma dual-equipment configuration, and his eyes were on the phone screen.
"Your two good plays were exceptional. Your two bad plays were worse than your baseline." The observation was delivered to the phone, not to Arisu. "The pattern is binary — you're either significantly better or significantly worse, with nothing in between. That's not fatigue."
He noticed. Of course he noticed. Kenma doesn't miss patterns, and the pattern of Future Branches use is distinctive — one play of impossible accuracy followed by one play of impossible wrongness. No natural player has that distribution. Natural players have gradients. Good plays, decent plays, mediocre plays, bad plays. A distribution curve. What I showed today was a step function — either perfect or catastrophic.
And Kenma is the kind of person who sees step functions in other people's performance and files them as data points that require explanation.
"Trying something new with my reads," Arisu said. His throat felt dry. He drank water. "It's inconsistent."
"Mm." The sound that was never just a sound. Kenma's thumbs moved on his phone. "The inconsistency has a pattern. Your good reads came with a delay — about one second before the play where you paused, then committed fully. The bad reads had the same delay and the same full commitment. You're not guessing. You're seeing something, and sometimes you're seeing wrong."
He timed the pause. He identified the commitment pattern. He isolated the variable — the one-second delay before the prediction resolves — and connected it to the outcome quality. Kenma is modeling my ability without knowing what my ability is, and his model is getting dangerously close to accurate.
"I'll work on it." Arisu rubbed his temples again. The headache was persistent — not debilitating, but present enough that the court looked slightly too bright and sounds arrived slightly too sharp.
Kenma pocketed his phone and stood up. "You should work on making the delay shorter. If you're going to commit that hard, the data you're using needs to arrive faster."
He walked away.
Arisu sat on the bench with the headache and the water bottle and the memory of Kenma's analysis running through his mind on a loop. The observation was precise. The advice was sound. And the underlying implication — that Kenma had identified a mechanical pattern in Arisu's performance that no amount of "trying new reads" could explain — was another data point added to a file that grew thicker with every match.
Kenma's file on me now includes: impossible scouting report sources, a twenty-eight-minute performance wall, and a binary read pattern with a one-second pre-commitment delay. Each point is individually explainable — fatigue, new technique, experimental approach. Together they form a picture of someone who has ACCESS to information that normal players don't, and who LOSES that access at predictable intervals.
He's not suspicious. He's curious. The difference matters — suspicious people accuse, curious people collect. But curious people collect until they have enough data to form a conclusion, and Kenma's data set is growing.
Misaki Residence — 11:47 PM
The dream interface opened to the infinite court. Arisu stood at center position and replayed the four Future Branches activations from practice, each one preserved in the system's memory like video files waiting for review.
Activation one: Kuroo's cross. Prediction: cross. Result: correct. He analyzed the decision tree — what had made this prediction accurate? Kuroo's approach had been committed early, the arm angle definitive, the trajectory locked before the ghost-image generated. The prediction succeeded because the target had already decided.
Activation two: Yamamoto's line-gone-cross. Prediction: line. Result: wrong. Yamamoto had been deciding between line and cross throughout his approach, and the prediction had been generated during the decision window — before the target committed. The sixty-percent model captured the most likely outcome at the moment of generation, but Yamamoto had chosen the less likely option after the prediction was already displayed.
The wrong reads came when the target changed their mind mid-motion. The prediction captures the most probable trajectory at the moment of activation, but it can't account for decisions the hitter makes AFTER the prediction generates. Against hitters who commit early — like Kuroo, like most power hitters — the accuracy rate is higher. Against hitters who decide late — like Yamamoto's read-hitting, like any smart attacker — the accuracy drops.
The freak quick is the worst possible case for Future Branches. Hinata's approach gives no directional information because his eyes are closed and his body language is committed to everywhere and nowhere simultaneously. The prediction would be capturing Kageyama's set delivery, not Hinata's decision, and Kageyama's precision means the set goes to the same spot every time — the SPIKE direction is determined by Hinata's hand angle at contact, a decision made in the last quarter-second of the approach, after any prediction window.
Future Branches can't solve the freak quick. The ability that I chased for weeks and unlocked in the middle of a match against Karasuno is useless against the one attack that matters most.
But it's not useless. Against standard hitters — Tanaka's committed crosses, Asahi's readable power shots, any spiker who decides direction during their approach — sixty percent accuracy on critical rallies is a significant edge. The tool doesn't solve every problem. It solves specific problems under specific conditions at a specific cost.
Scalpel. Not sledgehammer. Same lesson, bigger scale.
He dismissed the replays and pulled up the dream interface's training calendar. The Inter-High preparation period was starting — Kuroo had mentioned practice matches forming for the next two weeks. Three opponents. Mixed canon data.
Three practice matches before the tournament that starts Nekoma's path toward Nationals. Three chances to test Future Branches in competitive conditions. Three chances to build the use-case database — when to activate, when to conserve, which hitters the prediction model handles well and which ones break it.
And somewhere in those three matches, the physical fundamentals need to keep climbing. Receiving at forty-two percent isn't enough. Blocking at twenty percent isn't enough. The system augments skill — it doesn't replace it, and every match I play without MS proves that the player underneath the system is still a work in progress.
The dream interface shifted. A new screen, unbidden — the practice match schedule that Kuroo would announce tomorrow morning. Three matches. Two weeks. Opponents Arisu had mixed canon data on.
He closed the interface. The headache from practice had faded to a memory. The body was exhausted but the mind was awake, running calculations on activation timing and prediction accuracy and the shrinking margin between what Kenma suspected and what Kenma could prove.
Three matches. Two weeks. And Kenma is watching.
Want more? The story continues on Patreon!
If you can't wait for the weekly release, you can grab +10, +15, or +20 chapters ahead of time on my Patreon page. Your support helps me keep this System running!
Read ahead here: [ patreon.com/system_enjoyer ]