Late night in Lujiazui. The financial tide had temporarily receded, leaving behind a digital beach awaiting the next tide's inscription. Mozi's trading room was immersed in near‑absolute silence, broken only by the low, constant hum of server‑cluster cooling fans—like the breath of a slumbering beast. On the huge main screen, market‑data flows had slowed, replaced by complex model‑structure diagrams and performance‑analysis reports.
His [Trend Model], after weeks of intense development and multiple rounds of historical‑data back‑testing, had taken initial shape. The model integrated several indicators—RSI, moving averages, ADX—trying to launch a guiding flare over the battlefield where bull and bear forces intertwined. The back‑test curve shone brightly in most market phases with clear trends, effectively capturing and following the main movement direction, proving the basic framework's feasibility.
Yet Mozi's deep gaze did not linger on those satisfying profit‑making stretches. Instead, it locked onto the glaring "drawdown" intervals and "choppy‑market misjudgment" marks in the back‑test report. In these regions where market states were ambiguous and the boundary between trend and noise blurred, the [Trend Model] appeared hesitant, even clumsy. It would close trend positions prematurely because of a brief RSI divergence, missing larger subsequent profits; it would also hesitate to enter when a trend was just budding and signals still weak, due to contradictory signals from other indicators, losing the first‑mover advantage.
The root of the problem lay in the fact that the [Trend Model] was essentially a collection of **heuristic algorithms** based on statistical regularities and threshold judgments.
**Heuristic algorithms**, as the name implies, rely on "rules of thumb," "intuition," or "tentative strategies" to find satisfactory (though not necessarily optimal) solutions. They are widely used in financial modeling because markets are too complex to be fully described by precise mathematical models. Like an experienced old sailor judging weather trends by observing clouds, wind direction, and waves—not always 100% accurate, but making effective decisions most of the time.
The rules in Mozi's model—such as "be alert for trend reversal when RSI crosses above 70 and ADX is greater than 25," "confirm trend establishment when price breaks through the 60‑day moving average with volume expansion"—were typical heuristic rules. They were flexible, able to adapt to market nonlinearity, but for that very reason were full of **uncertainty** and **ambiguity**. These rules were like a group of staff officers each offering advice from their limited perspective (a single indicator), lacking a supreme commander to make a final ruling based on more fundamental principles.
This caused the model, when facing complex, contradictory market signals, to easily fall into "internal conflict" and "indecision." Its decision‑making process resembled a black box, lacking solid logical grounding and explainability. The model's "soul" seemed scattered, empiricist, even somewhat "opportunistic."
Mozi longed to imbue his model with a more stable, profound "soul." He needed a logical kernel that could pierce through market short‑term noise and touch deeper‑layer order. He thought of Yue'er, of the mathematical world she studied—a realm pursuing absolute rigor and necessity.
He reopened a recent paper by Yue'er. This paper wasn't directly about the P/NP problem, but explored the profound connection between "determinism" and "constructibility" under specific algebraic structures. In it, Yue'er rigorously argued that on mathematical objects satisfying certain symmetry and smoothness conditions (e.g., certain types of algebraic varieties), there existed an "intrinsic determinism"—some of their key properties (like genus, singularity type) were unique and inevitable, computable from their defining equations through a series of definite, finite steps (**deterministic algorithms**), independent of probability, independent of random probing.
**Deterministic algorithms**, their core characteristic being that for the same input, executing the same steps must yield exactly the same, uniquely determined output. They contain no randomness or ambiguous judgments. Solving the roots of a quadratic equation is a simple deterministic‑algorithm process.
Yue'er emphasized in the paper that this "intrinsic determinism of mathematical objects" stemmed from the "rigidity" and "well‑definedness" of their underlying structure. This determinism was the cornerstone of mathematical truth's solidity.
The paper shone like a strong beam into Mozi's contemplation of his model's "soul." Financial markets certainly lacked the perfect, static determinism of mathematical objects. But could there be some deeper‑layer, relatively stable "market structure" or "state," whose key features could be identified and judged **deterministically**?
A bold idea took shape: Could he abstract and transform the thought about "mathematical‑structure determinism" from Yue'er's paper and introduce it into his [Trend Model] as a higher‑level "state arbiter"?
He wasn't trying to directly use mathematical theorems to predict prices—that was impossible. He wanted to borrow that **mindset of pursuing deterministic judgment**.
He began redesigning the model's top‑level architecture. In the new framework, the various underlying heuristic indicators (RSI, MA, ADX…) would still act as "sensory organs," collecting signals from different market dimensions. But they would no longer directly vote or weighted‑average to decide trades.
Instead, a newly added core component, called the **[Market‑State Deterministic‑Identification Module]**, took center stage. This module's inspiration came directly from Yue'er's paper's exposition on "structural determinism."
The module's task: comprehensively analyze all signals provided by the underlying indicators, but its goal wasn't to predict prices. It tried to **deterministically** answer a higher‑level question:
* **Is the current market in a state with an intrinsically "trend‑friendly" structure?**
* This judgment would be based on a set of strictly defined, "rigid" logical conditions.
* For example, it might require simultaneous satisfaction: long‑term moving‑average groups showing clear bullish/bearish alignment (structural stability), ADX consistently above a threshold indicating trend strength (trend‑momentum confirmation), volatility within a moderate "trend‑channel" range (excluding extreme noise and panic‑driven choppiness)… etc.
* These conditions had to be clear, strictly testable, like premises in a mathematical theorem. Only when **all** conditions were simultaneously satisfied would the module **deterministically** output: "Current market is in an identifiable trend‑structure state."
Once this "deterministic" judgment was made, it would become the supreme command. Only then would the model allow the underlying heuristic rules of the [Trend Model] to operate freely within this "certified" trend environment, executing specific entry, position‑adding, stop‑loss actions.
Conversely, if the [Market‑State Deterministic‑Identification Module] could not reach the deterministic conclusion that "trend‑structure state holds" (i.e., any one rigid condition unsatisfied), then no matter how tempting the signals from underlying RSI, MACD etc., the model would be forced to stay on the sidelines or run with minimal positions, avoiding risk in ambiguous market environments.
It was like, before dispatching an army (trend strategy) into an area, sending a reconnaissance team (deterministic‑identification module) to strictly evaluate, according to a checklist (rigid conditions), whether the area was **definitely** suitable for large‑force deployment. Only upon a definite "yes" would the main force enter. Otherwise, even if some uncertain opportunities were missed, it would preserve strength and avoid getting bogged down.
This architecture fused the flexibility of **heuristic algorithms** (underlying tactics) with the strictness of **deterministic algorithms** (top‑level strategy). Heuristic rules were responsible for capturing opportunities within the "deterministic"‑delineated safe zone; deterministic judgment provided the whole model with a "philosophical foundation" and "disciplinary soul," greatly reducing irrational decisions in ambiguous zones arising from heuristic‑rule contradictions or over‑fitting.
Mozi immersed himself in this architectural‑reconstruction work, feeling an unprecedented intellectual satisfaction and delight. This wasn't merely technical optimization; it was more like the infusion of a philosophical idea. Yue'er's mathematical thought—that obsession with "determinism" and "structure"—had crossed disciplinary divides, offering crucial inspiration for solving his own domain's problems.
In this process, his perception of Yue'er quietly shifted. Initially, she was an interesting academic interlocutor offering a unique perspective. Later, the New York dinner let him witness the depth and clarity of her thinking. Now, her work could so directly, profoundly influence the very "soul" design of his core model.
This influence wasn't simple knowledge transfer, but a paradigm‑level enlightenment and elevation. What he admired wasn't just the mathematical knowledge she possessed, but her extraordinary wisdom that penetrated appearances and reached structural essence, and the nearly artistic creativity and rigor she displayed while constructing edifices in the purely rational world.
He found himself thinking of her from time to time—the expression on her face when she focused on explaining "varieties," the light shimmering in her eyes when she answered "whether truth is computable"—a mixture of humility and firmness. These recollections carried an intellectual attraction and a subtle emotional resonance.
He paused coding and sent Yue'er an email. This time he didn't discuss specific mathematical problems, but shared some core insights from his process of integrating "deterministic thinking" into the model architecture, writing sincerely:
"…Your paper's discussion on 'structural determinism' was like a distant lighthouse keeper unexpectedly casting a steady, clear beam while I explored the model's soul in the fog of market data. It made me realize that even in the most chaotic systems, seeking local and temporary 'deterministic cornerstones' may be more fundamental than chasing global, vague 'probabilistic advantages.' Thank you deeply for the inspiration your work brought—far beyond the scope of academic exchange."
After sending the email, Mozi leaned back in his chair, gazing out at Shanghai's never‑sleeping, dazzling nightscape. His model was still under construction, far from perfect. But this time, he felt the model's "soul" had become clearer and more solid. That soul not only sprang from his insight into market patterns and code logic, but also quietly incorporated a strand of deep understanding—about determinism and structure—from a distant mathematical realm, and a growing admiration and appreciation for the intelligent woman who had enlightened him.
He knew the connection with Yue'er had already transcended simple cross‑disciplinary interest, beginning to penetrate deeper into the core of his career and personal emotions. This change was soundless yet carried undeniable force, like an ocean current deep beneath the surface, quietly altering the course of his thoughts and feelings. The night was deep, the world of code remained cold and rational, but the source of inspiration injected into it brought an unexpected warmth and light to his solitary domain.