Alpha essentially comes from 'pricing deviations'.
Table of Contents
1. Momentum Strategy: Trends are a risk premium in themselves.
2. Mean Reversion: The market corrects itself, but it won't save trash assets. 3. Statistical Arbitrage: Quant trading has truly entered the 'industrial age'.
4. The AI Era: Quant trading is evolving from 'strategy competition' to 'system competition'.
5. The unified framework of the three major strategies: they are fundamentally not contradictory.
6. True institutional-level quant isn't about 'predicting the market', but about managing risk.
VII. Conclusion
Introduction
In the previous article, we discussed the core underlying logic of quantitative trading:
Markets are not perfectly efficient; alpha essentially comes from "pricing deviations"; and quantitative trading uses systematic, probabilistic, and industrialized methods to continuously capture these deviations.
But the real problem is:
What exactly does an Alpha in the market look like?
The answer is not complicated.
Over the past two centuries, whether in stocks, futures, foreign exchange, bonds, commodities, or today's crypto asset market, the core strategies that have truly proven effective over the long term and have been continuously iterated upon by top global quantitative institutions have always revolved around three main frameworks:
Momentum strategy
Mean Reversion Strategy
Statistical Arbitrage Strategy
These three types of strategies essentially constitute the core "Alpha industrial system" of the modern quantitative asset management industry.
These are not historical sites found in textbooks.
On the contrary:
Today, the world's largest quantitative firms—including Renaissance Technologies, AQR Capital Management, and D. E. Shaw & Co.—are still running advanced variations of these strategies.
The real difference never lies in "knowing the strategy name," but rather in:
Who truly understands the market structure, behavioral finance logic, risk exposure, capacity limits, and strategy failure mechanisms behind them?
because:
All quantitative strategies are essentially "probabilistic machines that are effective in specific market environments".
I. Momentum Strategy: Trends themselves are a form of risk premium.
What is momentum?
The core logic of the momentum strategy is extremely simple:
Assets that have performed strongly in the past tend to remain strong in the future; conversely, assets that have been consistently weak in the past tend to remain weak.
This sounds like "buying high and selling low".
But true institutional momentum isn't about emotionally chasing highs, but rather:
Systematic arbitrage based on the inconsistent speed of information dissemination in the market.
Why can momentum persist for 200 years?
In 1993, Narasimhan Jegadeesh and Sheridan Titman published a classic paper:
(Returns to Buying Winners and Selling Losers)。
They discovered:
Buy the stocks with the strongest gains over the past 3-12 months, and short sell the stocks with the biggest losses over the past 3-12 months.
Even after deducting transaction costs, significant excess returns can still be obtained in the long run.
Subsequently, numerous studies worldwide further validated this:
The momentum effect exists not only in:
US stocks
European stocks
Commodity Market
Foreign exchange market
bond market
Crypto asset market
It even exists in data from the 1800s.
This means:
Momentum is not a random phenomenon of a particular era, but a long-term characteristic of the structure of human financial behavior.
The essence of momentum: the market's "slow reaction" to information.
Behavioral finance: Humans are not rational machines
Behavioral finance believes that:
Most market participants do not complete "full pricing" the moment information appears.
Conversely, humans will exhibit the following:
Underreaction
Herding
Confirmation Bias
Disposition effect
For example:
A company's performance exceeded expectations.
The market may only rise by 5% on the first day.
However, as more funds, institutions, and analysts gradually confirm the logic, funds will continue to flow in over the next few weeks or even months.
then:
Prices form trends.
That's momentum.
The "slow reaction" of institutional capital
Large organizations in the real world are not high-frequency robots.
Pension funds, sovereign wealth funds, insurance asset management, and bank asset management.
Their investment decisions often involve:
Approval process
Risk control restrictions
Position constraints
Liquidity constraints
therefore:
Large amounts of capital naturally lead to a "delay in price discovery".
Momentum strategies are essentially about arbitrage in this "slow pricing" scenario.
Risk Compensation: Momentum profits from "tail risk".
Another school of thought holds that:
Momentum works not because the market is inefficient, but because:
Momentum strategies inherently carry a significant risk of tail reversal.
in other words:
Momentum is not a risk-free arbitrage opportunity.
It is making money:
The risk premium of "trend continuation".
The greatest enemy of momentum: Momentum Crash
The most dangerous aspect of momentum strategies is not the slow and steady loss of money.
Instead:
During extreme reversals, a sudden collapse occurs.
The most classic case:
V-shaped recovery following the 2009 financial crisis
After the 2008 crash:
Previously, the worst-performing junk stocks suddenly rebounded violently, while the previously resilient quality stocks underperformed.
result:
Be a winner when you're bullish, be a loser when you're bearish.
The classic momentum combination resulted in a "double kill" in a short period of time.
Many strategies have resulted in losses exceeding 40% in a single month.
Momentum Stomp in the Crypto Market
In the crypto market, this phenomenon is even more extreme.
because:
Higher leverage
Liquidity is more fragile
Retail investors account for a higher proportion
More intense emotional fluctuations
A short squeeze following extreme panic often leads to:
Momentum strategies can retrace months' worth of profits within hours.
therefore:
The core of true institutional-level momentum trading is never "finding trends".
Instead:
How to survive a trend reversal.
II. Mean Reversion: The market will correct itself, but it won't save junk assets.
What is mean regression?
The logic of mean reversion is completely opposite to that of momentum:
When prices deviate from the "reasonable mean", they will eventually revert to the mean.
Simply put:
Too fast a rise → may fall back
A sharp drop may lead to a rebound.
This is one of the oldest ideas in the financial markets.
However, mean reversion is most easily misunderstood.
The biggest misconception among many retail investors is:
"A significant drop = a valuable asset"
This is incorrect.
For true mean reversion to hold, one prerequisite must be met:
"The mean itself did not collapse."
What assets won't be returned to their original value?
if:
Business model disappears
Industry eliminated
Liquidity Depletion
Credit system collapse
So:
A price drop is not necessarily a "deviation from the mean".
Instead:
The mean itself shifts permanently downward.
This is why:
Kodak
Nokia
Earth Moon
After the crash, there was no true "mean reversion".
What exactly is the money made from mean reversion?
Essentially:
They make money from "market overreaction".
For example:
A company's stock price plummeted by 20% due to short-term negative news.
but:
Core business remains unchanged
Cash flow remained unchanged
The industry logic remains unchanged.
So:
Market sentiment often leads to overpricing of risks.
The mean reversion strategy is essentially a gamble:
Emotions can be corrected.
The most classic example of mean reversion: Pairs Trading
Pair trading is one of the most classic mean reversion models in the history of modern quantitative trading.
For example:
The Coca-Cola Company vs PepsiCo
Bank stocks vs. bank stocks
Oil stocks vs. oil stocks
The logic is very simple:
If two companies are highly correlated in the long term, but their price discrepancies widen frequently in the short term,
but:
Go long on the undervalued side
Shorting the overvalued side
Waiting for the price difference to converge.
The real risk of pair trading: the price difference may never revert to its previous form.
This is the part that many novice quantitative traders easily overlook.
During the 2007 global "Quant Quake" (quantitative seismic event):
Many pair trading strategies failed simultaneously.
The reason is not a model error.
Instead:
Quantitative funds in the market use overly similar models.
When some funds are forced to deleverage:
The simultaneous closing of all positions caused the price difference, which should have converged, to widen further.
This exposes a harsh reality:
One of the biggest systemic risks in quantification.
It's called "strategic crowding".
III. Statistical Arbitrage (Stat Arb): Quantitative Trading Truly Enters the "Industrial Age"
If we say:
Momentum = Trend Arbitrage
Mean reversion = deviation arbitrage
So:
Statistical arbitrage, on the other hand, industrializes all alpha factors.
What is statistical arbitrage?
The core idea of statistical arbitrage is:
Instead of relying on a single big opportunity, it establishes a stable long-term return curve by leveraging a large number of low-probability advantages.
It's more like a casino.
Casinos don't need to win money every time.
It only requires:
There is a slight probability advantage in each round.
The number of bets is large enough
final:
The law of large numbers will take over the returns.
The same applies to statistical arbitrage.
The core process of statistical arbitrage
Step 1: Alpha Factor Modeling
Develop a multi-factor model for thousands of stocks:
Common factors include:
momentum factor
Value Factor
Volatility factor
quality factor
Liquidity factor
Market capitalization factor
Essentially:
It is predicting "future short-term excess returns".
Step 2: Portfolio Optimization
This is where an organization's capabilities are truly demonstrated.
A portfolio not only needs to make money, but also needs to:
Market Neutral
Industry Neutral
neutral style
Beta Neutral
Volatility is controllable
Controllable drawdown
therefore:
Modern statistical arbitrage is essentially more like:
"Risk Engineering".
Step 3: Execution System
Many people believe:
The most important aspect of quantification is the model.
In fact:
For large organizations
What truly determines life and death is often:
Slip point control
Trading shock
Liquidity Management
Order routing
Latency optimization
because:
Alpha is very thin, but its cost is extremely high.
Why is statistical arbitrage becoming increasingly difficult?
In the 1990s, statistical arbitrage was considered a "dividend from information asymmetry".
But today:
Data is becoming more transparent
Computing power is getting cheaper and cheaper
AI modeling is becoming increasingly popular
Factors are severely crowded
then:
Alpha continues to decline.
The industry is beginning to enter:
"Infrastructure arms race".
The competitive dimension is from:
Who knows the factor?
Gradually evolved into:
Who can calculate faster?
Who has lower transaction costs
Who has stronger risk management?
Who can continuously discover new Alphas?
IV. The AI Era: Quantitative Analyses are Evolving from "Strategic Competition" to "Systemic Competition"
This is the most significant industry change in 2025-2026.
past:
Quantitative analysis is about:
factor
Model
math
today:
AI is changing the entire industry structure.
What has AI quantification truly changed?
Alpha discovery automation
past:
Researchers need to manually identify the factors.
today:
AI can do it automatically:
Feature mining
Generating factors
Test Relevance
Perform out-of-sample validation
Discovering nonlinear relationships
Alpha production began to be industrialized.
Unstructured data enters the market
Traditional quantitative methods mainly rely on:
K-line
Financial Report
Trading volume
But after the emergence of AI:
The market has begun to process:
news
social media
On-chain data
Satellite data
voice
video
Macro text
This means:
AI is transforming "information gap" into "cognitive gap".
Intelligent risk management
The greatest value of AI may not be "increasing profits".
Instead:
Reduce the probability of strategy failure.
include:
Dynamic position adjustment
Real-time volatility identification
Liquidity risk monitoring
Regime Shift Recognition
True AI-driven quantitative analysis is not about predicting the market.
Instead:
Adapt dynamically to the market.
V. The unified framework of the three strategies: They are not inherently contradictory.
Many people believe:
Momentum and mean reversion are contradictory.
Actually, no.
They are simply:
Different manifestations of market efficiency deviation at different time scales.
Market Structure at Different Time Scales
Time Scale, Dominant Logic, Core Phenomenon: Ultra-short-term (intraday): Microstructure: Liquidity, market-making behavior; Short-term (1-12 months): Momentum: Incomplete information diffusion; Medium-term (1-3 years): Mixed phase: Coexistence of trends and corrections; Long-term (3-5 years+): Mean reversion: Valuation repair and overreaction correction. Statistical arbitrage:
This involves industrializing and integrating all these logics.
VI. True institutional-level quantitative analysis is not about "predicting the market," but about managing risk.
The biggest misconception in the quantitative industry is:
"Quantitative analysis = a high-win-rate money-making machine".
In fact:
Truly top-tier asset management institutions are never primarily concerned with:
Rate of return.
Instead:
Stability of risk-adjusted returns.
because:
High returns are unsustainable
High leverage inevitably leads to fragility
Alpha will decay
Market structure will change
Organizations that truly survive in the long run do not rely on being "always right".
Instead:
It is not fatal even when there are mistakes.
VII. Conclusion
Momentum, mean reversion, statistical arbitrage,
Essentially, these constitute the three core "Alpha matrices" of the modern quantitative world.
They are able to traverse 200 years of market history.
It's not because markets are always inefficient.
Rather, it's because:
Human nature, liquidity, risk appetite, institutional constraints,
It has never really changed.
But today, the industry has entered a new phase:
In the past, the competition was about "strategy"; now, it's about "system".
In the past, the competition was about "factors"; now, the competition is about:
AI capabilities
Data capabilities
Risk control capability
computing power
Execution capability
Organizational industrialization capabilities
The true era of AI-driven quantitative asset management
It is no longer:
"Who can write a strategy?"
Instead:
Who can continuously build an Alpha factory?
—AI-powered customized trading brand solutions for entrepreneurs, communities, crypto asset management institutions, and trading KOLs
Bubble VCBLO1