It was January 2010. Snow in London.
Julian arrived at the office early, as always. He took the farthest seat in the support section, hung up his coat, and powered on the terminal.
He had only intended to review a backtest model. But then a notification popped up—something outside his folder permissions:
[X-FinPrep] Q4 Earnings Template – PRELIMINARY
Last modified: Today, 5:17 a.m.
It was a draft—an internal earnings estimate from Risk Compliance. Technically, he wasn't supposed to see it. These documents were subject to access tiers and were usually locked until 9:00 a.m.
He didn't open the file.
But on the side panel, a live-edit log was visible:
EPS estimate updated: 1.41 → 1.68
A 20% upward revision in earnings per share.
If it was real, the market would explode.
If it was false, it would just be noise.
Julian stared at the number for two seconds.
Then switched windows.
He didn't place a direct trade.
What he built was a "structured basket"—a modelled trade that sounded complex but boiled down to this:
"Group a few stocks, identify correlated directional bias, simulate an entry."
It looked legal. It looked intelligent.
But he knew perfectly well:
He wasn't betting on fundamentals.
He was betting on this question:
"Has the market already sniffed something?"
This kind of move had a name—pre-event positioning.
Even in simulation environments, most firms didn't allow this.
You couldn't use unconfirmed data—even in a test.
Julian built the basket using three superficial factors:
"price momentum," "volume skew," and "sentiment pivot."
It was a beautiful lie.
Three technical terms, all perfectly aligned to suggest he had no idea about the EPS update.
He even made sure the ratios looked like they came from a textbook.
But the only variable that truly mattered was the EPS estimate.
At 9:26 a.m., he hit "run."
The simulator began projecting: if the stock showed strength today, then based on current volume distribution, it was likely to gain over 10% within three days.
The PnL curve rose steadily.
The system auto-flagged the take-profit level at +12.1%.
He pressed confirm.
The trade was logged.
He got up and went to the restroom to wash his face.
When he returned, there was a new yellow line on his screen:
RISK SENSITIVITY: ELEVATED
Behavioral monitoring activated.
Julian didn't flinch. Didn't smile.
He simply took out his notebook and wrote:
"The system is watching me."
"Good."
That evening, the company released its earnings guidance.
EPS: 1.68—exact match.
The stock rose 11.2% that day.
By the next close: +17.8%.
Julian didn't report it.
He reopened the simulator log and added three lines at the bottom:
"Might not be insider info."
"But the market smelled it early."
"I just tested the nose."
He walked home under a dark, post-snow sky.
On the way, he bought a cheap beer and a pack of salted crackers.
Not because he wanted them.
But to prove to himself that he could still buy junk like everyone else.
He drank three sips. It tasted like water.
He was satisfied.
That kind of tasteless, soundless, traceless experience—
That was exactly what the trade had been.
Clean. Cold. With just the right amount of guilt.
Of course he knew he'd crossed a line.
But there was a voice in his head that said:
"This world was never clean.
I just admitted it before you did."
He didn't go straight home that night.
He walked to the basement level of Selfridges, to that unnamed Italian cocktail bar with no signage. He didn't remove his coat at the door.
He liked places like this—no menus, no music, no small talk. No one asking if you had a good day.
He ordered a Negroni.
Not because he liked the taste—he hated orange peel oil.
But he liked its bitterness. Its structure.
Three spirits, equal parts. No improvisation. No syrup. No metaphor.
Just 1:1:1.
He didn't sit. Just stood at the bar, tracing the condensation on the glass rim with his finger—over and over.
The bartender added the ice cubes a little too slowly.
With each one, Julian's brain auto-projected an intraday market snapshot—
not the weight of the ice, but the layered buy walls behind today's stock.
The first cube: 9:26 a.m., when he triggered the basket.
The second: 11:04 a.m., when institutional volume broke the top wick.
The third: late afternoon—shrinking volume, a doji bar, traders mistaking silence for an ending.
He took a sip.
The bitterness had thinned. Slightly diluted.
Like a mistake—but not fatal.
He was pleased.
He drank slowly.
Every sip like a single cycle in a backtest.
He didn't blink. Didn't talk.
The bartender glanced at him once, then looked away and kept wiping the same glass.
Not very clean. That irritated him.
He hadn't brought his folder.
But he had printed three copies of the trade summary.
Stored them in a waterproof envelope inside his briefcase.
Labeled with two tags:
"Event-Based Driver Testing"
"Non-Correlated Synthetic Exposure"
Not for archiving.
Just to convince himself this was all "within system boundaries."
He stood a while longer, staring at his own reflection in the bar's glass display.
His face was blank.
His eyes were clear.
He looked like a decent man.
He whispered:
"I didn't buy any stock."
"Just observed a potential volatility vector."
When he got home, he didn't turn on the lights.
He took out the documents and taped the trade printout to the side of his fridge.
Right next to the one from last year's simulation blow-up.
Both charts were perfectly aligned.
Similar slopes, matched gridlines, LaTeX formatting, margins equalized to the millimeter.
He stood there for ten full minutes.
Then unplugged the refrigerator.
Not to save energy.
He just hated the buzzing sound.