In Lujiazui's trading room, time flowed in milliseconds, split into countless slices formed by market data. Mozi's [Oscillation Model], having survived the false‑breakout scare and undergone parameter optimization, ran increasingly steadily, like an experienced diver adeptly gathering pearls (tiny profits) in the market's shallows (narrow fluctuations). The capital snowball, driven by near‑silent compound‑interest effect, expanded continuously and stably; the slope of the account curve maintained a satisfying angle.
But Mozi understood well that the market's ocean wasn't only tranquil shallows. Undercurrents of the global macroeconomy were stirring—subtle shifts in major‑economy central‑bank policies, faint restructurings of geopolitical landscapes. These deep forces, like distant monsoons, would eventually stir huge trending waves. His [Oscillation Model], whose design philosophy was rooted in the statistical assumption of "mean reversion," excelled at capturing order amidst chaotic fluctuations, yet was inherently powerless—indeed, vulnerable—toward powerful, unidirectional "trends."
When a trend emerged, the market displayed a distinctly different "personality" from during oscillation periods. Price would form a series of higher highs and higher lows (uptrend), or lower highs and lower lows (downtrend), along one main direction (rising or falling). Attempting "buy‑low‑sell‑high" oscillation strategies in such an environment was like swimming against the current—not only missing the main profit band, but more likely being swallowed by continuous one‑way movement through constant "catching the bottom" or "touching the top," leading to huge losses.
The model's boundaries emerged clearly now. No model was omnipotent; its strength necessarily accompanied specific applicable scenarios and inherent blind spots. The boundary of the [Oscillation Model] lay precisely in its inability to recognize, nor adapt to, strongly trending markets. Like a system that can navigate precisely on land, once thrown into the ocean, it would utterly fail.
Mozi stood before the huge screen wall, gaze profound. He saw that the gold market's recent volatility was quietly increasing; price lingered in the fluctuation range for shorter times; occasional, slight breakouts, though still mostly "false moves," seemed to gain subtle strength and persistence. These were early signals that market state might be switching.
It was time. He needed to build a new "weapon," a sharp blade that could ride the waves in the ocean of trends—the [Trend Model].
He brought up a new development interface, began sketching the new model's framework. Unlike the [Oscillation Model] relying on gradient descent to find local optima (like the lowest point in a valley), the [Trend Model]'s core idea was identifying and following the market's main movement direction, capturing its "momentum."
To quantify market "momentum" and potential trend strength, a classic yet powerful technical indicator surfaced in his mind—the **Relative Strength Index (RSI)**.
Mozi didn't immediately start writing code, but habitually first constructed a clear conceptual model in his thoughts. The essence of RSI was a **momentum oscillator** measuring the total magnitude of price gains versus price losses over a period (typically 14 periods). It worked by comparing the average gain on up days versus the average loss on down days, assessing the relative strength of buying/selling forces.
He mentally simulated RSI's operation:
When the market was in a strong uptrend, up‑day magnitude and number often outweighed down days, causing the "average gain" in RSI calculation to far exceed "average loss," pushing RSI values toward the high region above **70**, even **80**. Conversely, in a strong downtrend, RSI values would slide toward the low region below **30**, even **20**.
This introduced RSI's two most classic application concepts—**overbought** and **oversold**.
"When RSI enters the region above 70, it's usually seen as an **overbought** state," Mozi reflected. "This doesn't mean the market will immediately reverse downward, but indicates that within the current rising momentum, buying power may have been relatively fully released; price relative to its intrinsic value might be over‑elevated; market sentiment tends toward frenzy; latent profit‑taking pressure accumulates. Like a compressed spring storing reverse‑motion potential energy."
"Similarly, when RSI enters the region below 30, it's seen as an **oversold** state," he continued reasoning. "This indicates selling power may have been excessively vented; market sentiment pessimistic to extremes; price possibly undervalued; latent bottom‑fishing buying possibly brewing. Like an overstretched rubber band possessing rebound potential."
But Mozi knew well that in trending markets, simply acting in reverse based on RSI entering overbought/oversold zones was extremely dangerous. Because in strong trends, RSI could linger in overbought zones (uptrend) or oversold zones (downtrend) for long times, even experiencing "obtuse" phenomena. Price would race along the trend direction, crushing those who "counter‑trended" too early.
Thus, in his [Trend Model] design, RSI's role wasn't a simple reversal indicator. He focused more on:
First, **RSI trend and divergence**. When price makes new highs, but RSI fails to simultaneously make new highs, forming "bearish divergence"—this might signal exhaustion of upward momentum, an early warning of possible trend reversal. Conversely, "bullish divergence" might signal exhaustion of downward momentum.
Second, **RSI breakout around the centerline (50)**. RSI breaking above 50 from below means bullish forces start dominating—could be a confirmation signal for an uptrend; breaking below 50 from above might mean bearish forces control the situation.
Third, **Combining RSI with price trendlines**. Drawing trendlines for RSI's own movement—its trendline break might give signals earlier than price's own trendline break.
He wove these thoughts into the new model's algorithmic logic. The [Trend Model] would be a multi‑factor decision system; RSI was its core momentum‑measuring component, combining moving averages (judging trend direction), Average Directional Index (ADX, judging trend strength), and other market‑breadth indicators to comprehensively judge whether the market was in an effective trending state, and issue trend‑following trading signals accordingly.
While constructing the model, a philosophical question spontaneously floated up in his mind: In the mathematical universe, did there exist clear, persistently running "trends" like in financial markets?
He recalled Yue'er, that scholar immersed in pure‑mathematics worlds. Her domain was full of eternal theorems, immutable axioms, and mathematical structures that seemed to exist timelessly. Euclidean geometry still held after two millennia; prime distribution patterns existed objectively, whether humans discovered them or not. Could this be seen as a kind of eternal "uptrend" in the mathematical world—an irreversible directional movement toward absolute truth?
On the other hand, mathematics' own development seemed filled with "oscillations" even "paradigm shifts." Discovery of non‑Euclidean geometry shook Euclidean geometry's absolute reign; Gödel's incompleteness theorems cast a shadow over mathematical certainty. This resembled "retracements" or "reversals" within trends.
His curiosity about Yue'er's research, about the mathematical‑foundation unity she pursued, grew stronger. What shape was that world's "order"? A static perfect sculpture, or a dynamic, evolving lifeform?
That curiosity drove him; after a day's work temporarily ended, stars already filling the sky outside, he again sent Yue'er an email. This time, the question was more abstract, touching more essential matters.
"Ms. Yue'er, recently building a new market model, thinking about the concept of 'trend.' In financial markets, trend means price's continuous directional movement over a specific period, driven by combined forces including intrinsic value, fund flows, market sentiment. It has a life cycle of birth, development, peak, decline, and is often disturbed by various 'noises.'
"This sparked a curiosity: In the mathematical universe you explore, does some kind of 'eternal trend' exist? For example, does mathematics' own knowledge‑system expansion exhibit an irreversible, upward 'trend'? Or, deeper within mathematical structure, is there some fundamental evolutionary direction pointing toward some ultimate goal? Or perhaps mathematical truth itself is static, non‑temporal, making the concept of 'trend' therein fundamentally inapplicable?
"Very interested in your view from a philosophy‑of‑mathematics perspective. This might help me better understand 'order's manifestations across different scales."
After clicking send, Mozi leaned back in his chair, gazing at the still‑jumping market data on the screen. His [Trend Model] was still just an embryo, needing vast historical‑data backtesting and parameter optimization before deployment. But he knew expanding models' boundaries, adapting to different market states, was evolution he must complete in this eternal game against market complexity.