109: Chapter 109 The Grand Vision of Quantification

The first tier of the tiered settlement, along with the base fee, totaled 300,000.

300,000, in this day and age where people casually talk about small targets of hundreds of millions online, hardly seemed like a large number.

But for Jiang Lin's family, it was definitely not small.

When buying groceries at the vegetable market, his mother could haggle back and forth with the vendors for several minutes over a price difference of thirty or fifty cents.

His father had thrift carved deep into his bones, capable of going an entire year without spending a single cent on extra clothing.

Yet for the things Jiang Lin planned to do in the future, 300,000 was far too small.

A few decent servers, several batches of hard drives, a set of high-precision sensor kits...

Tear it apart casually, and it would transform from a figure that made ordinary families' hearts skip a beat into a few rows of budgets quickly crossed off in a spreadsheet.

Jiang Lin opened his financial spreadsheet.

After recording this transaction, the number at the bottom of the spreadsheet jumped.

Since April, the ID 1453 had not been idle.

A small order with a misaligned reinstated price, 24,000.

A small order with a delayed suspended status field, 18,000.

A review of a misused financial report announcement date, 42,000.

A survival bias report for a stock pool, 36,000.

An instance where an old script version could not be reproduced, and the platform rushed to compensate 20,000.

And another field resampling anomaly, 31,000.

The amounts for each of these outsourced orders were not large.

But every clash between code and data was teaching the MPS engine to recognize a sly way data lied in the Real World.

At the foundation of finance, prices could change backwards due to dividends.

The Physics state of a suspension could be misaligned by a day in the database.

Listed companies' financial reports could appear like ghosts before the announcement was released.

In a seemingly fair stock pool, only those survivors who won until the end could remain.

And a seemingly insignificant line of code modification could make the old script never run back to yesterday's perfect return curve.

By the time today's massive sum of 300,000 was credited, his confirmed total income from quantification-related underlying technical services had quietly reached the milestone of over 500,000.

However, because the previous rounds of Wasteland preparation, workstations, hard drives, sensors, machining tools, and observation equipment continuously swallowed money, the liquid funds he could readily dispose of at hand were currently just over 300,000.

If he let these 300,000 quietly lie in his bank account and earn that meager demand deposit interest, it would eventually just turn into a string of dead numbers slowly worn thin by inflation and currency over-issuance.

Jiang Lin believed that he might be able to turn this money into a machine capable of self-reproduction in this dark forest composed of capital and data.

This thought had actually taken root and sprouted ever since he first came into contact with quantitative code.

Working as a data scavenger on the periphery of Shen Chengye's platform for so long—from initial anomaly detection, to auditing baseline statistic leakage, to cleaning up obsolete factor libraries for private equity, and to the feature pipeline optimization a few days ago that dragged down the entire private equity server room...

Jiang Lin could already be extremely certain of one thing.

His hands, this brain, and the mathematical Physics underlying architecture he built could see micro-dimensions in the underlying business of quantitative trading—which the vast majority of peers would never see in their lifetime.

His dimensional reduction strike on computational complexity and his cleanliness obsession with data authenticity were the rarest heavy weapons in this industry.

It was time to step into the field and try it himself.

However, Jiang Lin's mind was exceptionally clear.

He knew that being able to see at a glance where a code pipeline was secretly cheating oneself was completely different from holding real money and silver to gamble on tomorrow's jumping K-line after the market opened.

Going from processing underlying data for others to truly stepping into the flesh-and-blood live combat battlefield with his own money, separated in between was an entire cruel world he had never set foot in, filled with irrational emotions, black swan events, and liquidity exhaustion.

Therefore, even though his underlying technology already looked down on all heroes, in the matter of trading, everything had to start from scratch.

Moreover, there was a bottom line that could not tolerate the slightest concession.

It had to be one hundred percent his own funds.

He couldn't look for investors to raise even a cent, couldn't accept any client's discretionary asset management, couldn't establish a private equity fund that required filing, and couldn't have any organizational structure that required him to write net value reports or sign his name externally.

Only his own money.

The transfer in and out of funds circulated in a closed loop solely between his own bank account and securities account.

Strategy iterations, risk control alarms, and daily profit and loss figures were known only to him alone.

For him, the more silent this money-making machine was, the safer it was.

Safety overrode all returns.

Having made up his mind, Jiang Lin began to set up the stage.

He directly skipped the stock market.

A-share's rule system was full of thick high walls everywhere for retail investors who wanted to do programmatic trading.

The T+1 trading system directly locked down most of the high-frequency microstructures and mean-reversion logic he wanted to verify intraday.

The high stamp duty, the one-sided market that could not be easily shorted, and the strict restrictions of major brokerages on retail investors accessing quantitative API interfaces all made this place unsuitable to be his first experimental plot.

Although ETFs, convertible bonds, and margin trading also had their own gaps, for Jiang Lin's current capital volume and identity, winding into them would only drive up compliance, interface, and strategy complexity.

Finally, he cast his gaze towards the commodity futures market.

Using his ID card, he opened an account at a top-tier futures company.

Account opening, video verification, doing the tedious risk tolerance appropriateness assessment, signing and reading out that thick stack of risk disclosure statements in front of the camera.

After the account was opened, Jiang Lin originally thought the next step was to plug the API key into his own server.

As a result, the next email sent by the futures company's account manager pushed his plan back a notch.

Programmatic trading interfaces needed to be applied for separately.

Simulation environment test records, maximum single-pen lots, maximum daily commission counts, handling logic after abnormal disconnections, and whether it had automatic order cancellation and risk control circuit breakers all had to be filled into the form.

Jiang Lin spent a day running through market subscription, order placement, order cancellation, execution reports, disconnection reconnection, and fund queries in the simulation environment.

The next day, official interface permissions were opened.

Next came the construction of the infrastructure.

He pragmatically rented a low-latency cloud server at a domestic backbone network node.

The function of this server at the current stage was very plain.

Provide a stable network connection, a fixed-configuration Linux running environment, be able to run 7 * 24 hours without interruption, and completely save every tick data pushed by the market interface and every commission order log issued by the system without omission.

In the first phase of trading, what Jiang Lin wanted was not the ultimate speed that nothing under heaven could outrun.

It was not to suddenly drop offline during critical market trends, not to lose the underlying records needed for troubleshooting when systematic errors occurred, and not to let an accidental hibernation of the laptop or a forced reboot of the Windows system ruin a whole day of his painstaking verification.

As for the quantitative software stack, he did not fall into the geeky obsession of reinventing the wheel from scratch to hand-write the entire market receiving engine.

Instead, he decisively adopted an open-source lightweight trading framework verified by countless live trades, using it to connect to the futures company's CTP interface for communication.

And at the core strategy calculation layer, he smoothly stitched his self-developed MPS-TimeLine time-series database, as well as those window operators proven absolutely correct by the MPS-Kernel using the Zero-One Principle and squeezed to the limit on the underlying micro-architecture, seamlessly into the very bottom layer of data processing.

With this foundation, his feature calculation engine ran shockingly fast.

But Jiang Lin was crystal clear in his heart.

The underlying code running fast was only a victory of basic performance.

It was by no means his winning move in this bone-spitting capital market.

For quantitative trading, the real life-and-death tribulation lay in the two words: backtesting.

And Jiang Lin's first ironclad rule for backtesting was bought with others' blood and tears from that platform baseline where he personally exposed the statistical quantity mixing vulnerability.

Do not lie to yourself.

In this industry, he had seen too many backtest curves that were breathtakingly beautiful.

That platform's official baseline put the training set and test set together for normalization, letting future information quietly seep into the past, and the backtest curve was pretty enough to fool a room full of people.

Therefore, the rule Jiang Lin set for himself was that for the generation of any trading signal, when the simulation system reached a certain historical moment T, the data it could call and see could only be the data that truly existed and had already occurred at and before time T.

Even future figures of T+1 milliseconds were strictly forbidden from leaking into the current state machine's computing memory by even a fraction.

Aside from future functions, the commodity futures market also had a natural trap specifically designed to slaughter countless quantitative recruits.

The rollover of the main contract.

Because futures contracts had delivery periods, funds would massively migrate from old contracts to new contracts a few months later as the delivery month approached.

Therefore, today's contract with the highest trading volume in the market, known as the main contract, and the main contract from the beginning of last month were physically not the same contract at all.

Due to storage costs, seasonal supply and demand differences, and the existence of risk-free interest rates, there naturally existed a huge price gap between old and new contracts.

When code crudely spliced them together to form a continuous historical K-line, a huge gap of hundreds of points would inevitably appear on the chart.

Jiang Lin knew how many self-confident novices triggered fake super-profit signals on this splicing gap that did not actually exist in real trading, and ultimately died a horrific death in the mincer of live trading.

He spent two days personally writing a rigorous reinstatement and contract smooth transition algorithm to smooth out this trap.

After the basic pits were filled, what came next was the friction cost that truly tested human nature.

Trading was never solving equations on a vacuum blackboard.

With every trade, the exchange and the futures company would take away handling fees.

When you saw a favored price on the order book and wanted to eat it up, due to your network speed and system latency, the truly executed price would often be worse than what you saw by a few ticks—this was slippage.

And when your capital volume was slightly larger, a market order you smashed down would directly eat through the pending orders of the first few tiers, pushing up your own costs—this was impact cost.

The backtests of the vast majority of amateur retail investors and half-baked quantitative researchers were all making effortless profits in a perfect utopia without friction and resistance.

Once it came to live trading, these invisible micro-losses alone were enough to eat up the meager theoretical profits on the books clean like army ants, or even leave you owing a mountain of debt.

Jiang Lin did not show the slightest leniency.

Build tables for each variety separately: minimum price fluctuation, contract multiplier, exchange handling fee, futures company additional collection part, close-today difference, main contract average order book thickness, and abnormal slippage in the first three minutes before opening and the last five minutes before closing.

Then, transform all these possible friction losses, according to a pessimistic caliber, into various penalty factors added into the cost function of the backtest engine.

Bilateral highest-tier handling fees, at least one to two ticks of slippage penalty, liquidity discounts under extreme market conditions...

After adding all friction parameters, he pressed the backtest button again.

A few seconds later, the return curve on the screen, which had been fairly smooth upwards, was as if someone had given it a heavy, muffled blow on the head, collapsing downward by a large chunk visibly to the naked eye.

The slope became gentler, the receding pits grew deeper, and overall it looked like a rugged, dilapidated mountain path.

Looking at this incomparably ugly curve, Jiang Lin felt no frustration at all; instead, he let out a long breath of relief.

Being ugly meant it was right.

Ugly was the true face of this cruel market.

Ugly was what was real.

In fact, at this stage, as long as he was willing, relying on his mathematical intuition and the powerful computing power of the MPS engine, he could completely turn back and add dozens of complex non-linear parameters to the strategy library.

He could easily use genetic algorithms or deep learning to repeatedly fine-tune and brute-force search these dozens of parameters until that curve was fed back into something steep, smooth, and pretty.

But Jiang Lin knew better than anyone that this carefully fed pretty curve was only valid for the past historical data he used to tune the parameters.

This was like a tight-fitting garment tailored specifically for a particular size.

Once it was taken to tomorrow, which was full of unknown noise, for live trading, it would instantly break its lines and reveal its true colors due to slight deformations in the market.

That was not called an investment strategy at all.

Statistically speaking, that was called extreme overfitting of historical data.

It was merely an illusion draped in a high-tech cloak.

Therefore, among these countless gorgeous algorithm options, Jiang Lin ultimately chose a system whose logic was simple to the point of being naive.

The underlying logic of this strategy was clean, with only a mere two or three parameters.

It did not pursue catching every tiny fluctuation, nor did it expect to buy the bottom against the market trend during a crash.

But its resilience was extremely strong; across several completely different historical intervals sliced out by Jiang Lin—which included bull-bear transitions, black swan crashes, and prolonged sideways trading—although it performed somewhat stumblingly, it could steadily maintain a mathematical expectation advantage that was always positive.

On the fourth day, with everything ready, Jiang Lin officially connected to the API key of the live trading server.

He went online.

But the position he invested was pitifully small.

There were a full three hundred thousand lying in the account, but when starting the strategy for the first time, he only used a tiny fraction of that principal as margin.

For every opening position instruction issued by the system, the risk exposure behind it was tightly locked down by his rigorous underlying asset volatility model, never exposing it within the direct strike range of a black swan.

Not only that, in this system, he also wrote a self-destructive tendency, Physics circuit-breaker level drawdown line.

Once triggered, the entire automated trading system would be hard circuit-broken at the trading gateway layer, canceling all unexecuted orders, prohibiting new position openings, and retaining only position-closing instructions.

Stop trading, exit the market.

Never strike a second blow, never stubbornly double down on positions out of spite due to losses.

The current Jiang Lin, holding three hundred thousand in his hands, relying on his ability to earn outsourcing fees, could actually fully afford this bit of early trial-and-error cost.

Over the past few nights, Zhang Xiufen had come in several times carrying cut fruit, seeing her son spacing out once again in front of a full screen of numbers and curves.

"Son, it's already been several days since the Gaokao, why are you still cooped up in your room messing with the computer every day?"

Zhang Xiufen placed the fruit plate on the edge of the desk, her tone full of heartache and puzzlement.

"You are already a great mathematician praised even on the news; there's no need to push yourself so hard anymore. Look at Yaoyao, the second after she finished the exam, she dragged a few classmates to Dali for a trip to relax. You should also go out for a stroll, or find classmates to play ball."

"I'm organizing some underlying structural data, almost done, Mom." Jiang Lin didn't even raise his head, his hand naturally switching away those windows flickering with green and red.

Looking at her son's engrossed back, Zhang Xiufen could only sigh helplessly, gave a reminder to rest well and not ruin his eyes, and then closed the door and retreated outside.

In her cognition, she just assumed that Jiangs Brick, which had been going crazy all over the world recently, still had countless profound subsequent questions that needed her genius son to answer.

No one would connect the computer in a high school senior's room with the server in some financial data center hundreds of kilometers away that was receiving the exchange's market data stream.

That being said, on the first day of live trading, the ruthless market truly gave this fledgling genius youth a harsh reality check.

Even though Jiang Lin had set the friction costs very high in the backtest, the complexity of the Real World still exceeded the predictions of the static model.

For a few chasing orders based on momentum breakthroughs, the moment the system's market orders smashed toward the exchange, the weak liquidity hanging there instantly withdrew and vanished like a school of startled fish.

This caused the true average execution price of these few orders to be one to two tick points worse than the worst-case scenario he assumed in the backtest.

The liquidity of the real order book, under the repeated tearing of algorithmic capital and hot money, was far from being as gentle and abundant as it looked in the historical slice data.

On the first day's close, the system settled, losing a few hundred yuan.

Replaced by any ordinary trading novice, or even some inexperienced quantitative researchers, at this time they would most likely start panicking.

They would begin to doubt whether their strategy logic had a fundamental error, frantically modify parameters after the market close, lengthen the period of the moving average slightly, or magnify the stop-loss amplitude slightly, attempting to erase today's few hundred yuan loss in the backtest.

Jiang Lin simply and calmly exported all dozens of real transaction records issued by the system on that day.

Then he cross-checked the timestamps, prices, and slippage of these dozens of transactions line by line and order by order with the simulated transaction orders generated by the local backtest engine at the exact same moments.

The ultimate conclusion reached was that the loss was not because the core trading strategy logic was wrong.

It was purely because in the cost model, his estimation of the slippage and liquidity exhaustion at the moment of the live breakthrough was not pessimistic enough.

After confirming this point, Jiang Lin did not move even a single hair in the strategy code library.

He simply opened the environment configuration file and turned up the slippage penalty parameter in the cost model by one notch in a more vicious and pessimistic direction.

To make the backtest closer to that maliciously full Real World.

On the second day, it was still fully automated machine operation, and Jiang Lin only acted as a bystander.

On the third day, the fourth day, the fifth day.

Five days data certainly could not prove the strategy was effective.

But at least one thing could be confirmed: the live trading did not show any systematic deviation visible at a glance.

Transaction direction, slippage distribution, commission consumption, order cancellation failure rate, and intraday equity fluctuation all fell within the pessimistic range provided by his local sandbox.

For the first week, this was already sufficient.

Because this was precisely the result Jiang Lin wanted to see most.

Every breath of the live trading matched the deduction of the backtest, which showed that from data cleaning, feature extraction, signal generation, all the way to the final order routing, he had not deceived himself in any tiny link.

At three o'clock on Friday afternoon, the selected set of daytime trading varieties ended the final trading session of the week.

Jiang Lin did not include the night session in the first-week prototype test.

The night session's liquidity, breaking news, and disconnection risks were more complex; those were variables for the next stage.

Thus ended the first week of trading.

He pulled up the backend liquidation and settlement statement and made a grand liquidation of the cash flow of these five trading days.

Deducting all commissions paid to the exchange and futures companies, every solid slippage loss generated by order book fluctuations, and that real money loss generated on the first day due to model error.

After deducting all Real World friction, the account made a net profit of over ten thousand yuan.

[Initial Capital: 300,000.00 Yuan]

[Current Equity: 310,842.63 Yuan]

[Net Profit: 10,842.63 Yuan]

[Weekly Return Rate: 3.61%]

[Maximum Drawdown: 1.18%]

A mere ten thousand yuan seemed insignificant compared to the tens of thousands of yuan in remuneration he got from catching a few system vulnerabilities.

Yet when Jiang Lin looked at these unremarkable numbers, the trembling sensation in his heart, intertwined with a sense of groundedness and excitement, was even several points stronger than when he had originally deduced a key lemma of the narrow theorem.

Because this was not just over ten thousand yuan.

It was even more a living system that he had built from scratch, line of code by line of code with his own hands, verified personally in the real flesh-and-blood grinding mill, and successfully withstood the extreme friction test of the real market in its first week.

It was indeed very small right now.

But as long as the process was rigorous, the data did not deceive him, the costs were not deliberately ignored by him, and the risk exposure was locked in a cage, this system possessed the qualification to continue iterating.

It was small right now simply because two things were currently both in the seedling stage.

The first was his principal.

This profit of just over ten thousand was not run out by going all-in with three hundred thousand at all, but rather the result of running out of a tiny pinch of edge principal within the three hundred thousand whose risk exposure he strictly controlled.

As long as this system could survive, as time passed, the gears of compound interest would begin to mesh, the snowball of the principal would roll larger and larger, and this absolute profit figure would naturally rise along with the tide.

The other was the depth of his strategy.

What he currently had running on the live market was merely a basic momentum system chosen to run through the entire underlying data link.

Because frankly speaking, when facing the real financial market, he still admitted that he was a novice in the trading field.

His understanding of the complex craft of market microstructure had only just pulled back its curtain.

But craftsmanship could be continuously polished through actual combat.

And the strongest trump card in his hand, the MPS architecture, was able to continuously iterate and upgrade as his mathematical research in the Wasteland and the Real World deepened.

He firmly believed that with every leap in the technical power of his underlying architecture and every enhancement in his squeezing of computing power, he would be able to find deeper non-efficiency regularities in this treacherous market, thereby squeezing out tenfold or even hundredfold cash flow from this ruthless machine.

Principal, strategy, underlying tool chain.

These three things would act like three giant interlocking turbines, accelerating and rolling upward together in the days to come.

Of course, Jiang Lin's mind was exceptionally clear-headed.

He knew that the small capital running out a weekly return of 3.61% right now did not mean that this proportion could still be maintained after the capital volume became large in the future.

When the capital volume expanded to a critical point, the massive orders themselves would become a giant crocodile alarming the market, and it would instantly trample the tiny arbitrage opportunities originally relied upon by the strategy into pieces.

By then, the impact cost would rise exponentially.

This still required him to continuously probe and test that invisible capacity boundary in future live trading.

Therefore, even though he had painted a grand blueprint in his mind that was enough to make people crazy, Jiang Lin still did not get carried away.

The market was a monster that could turn around and bite back at any time; any simple advantage that looked effective at present would rapidly decay as time went on and other smart money flooded in.

The bit of extra profit in the account today might very well evaporate into thin air by the morning of next Monday due to a sudden macroeconomic policy, or even cause him to lose his principal in reverse.

He withdrew his thoughts and opened an encrypted document named Trading_Journal on his desktop.

Every real transaction within this first week of live trading, every subtle deviation between the intraday trading and the backtest, and every notch of slippage assumption parameters he adjusted with his own hands were recorded word for word into the electronic journal.