Quantitative Trading: Seventy-Plus Years from Casino Math to the AI Arms Race
Research Date: May 2026 | Domain: Financial Investment | Subject Type: Industry/Concept Research Method: Horizontal-Vertical Analysis
I. One-Sentence Definition
Quantitative trading is a trading approach that uses mathematical models and computer programs to replace human intuition in making investment decisions. Its core belief can be summed up in one sentence: There are patterns in the market that can be captured with data, and machines are better at capturing them than humans.
II. Vertical Analysis: Seventy-Plus Years from Casino to Wall Street
2.1 Academic Seeds and Casino Experiments (1952-1973)
The story of quantitative trading has two threads: one in academia, the other in casinos and trading floors. They didn't truly converge until the 1980s.
The academic thread came first. In 1952, Harry Markowitz published Portfolio Theory, telling the world for the first time through mathematics: diversification is not intuition -- it's an optimization problem. In the 1960s, William Sharpe developed the CAPM model, and Eugene Fama proposed the Efficient Market Hypothesis. These theories may seem unrelated to "trading," but they provided the foundational language for quantitative trading -- risk can be quantified, returns can be priced, and markets can be modeled.
The practical thread started in casinos. In 1962, Edward Thorp published Beat the Dealer, proving mathematically that card counting in blackjack could yield a consistent edge. Casino owners were furious, but Thorp was really asking a bigger question: if math can beat the dealer, can it beat the market? In 1967, he published Beat the Market, answering with a resounding yes -- developing a convertible bond arbitrage strategy and later founding the Princeton-Newport Partners hedge fund.
Thorp's significance wasn't how much money he made -- it was the paradigm he pioneered: let data and models do the talking, not gut feelings and insider tips. Before him, investing was seen as a craft, something requiring "market instinct." Thorp said no: this is a game that can be dissected by mathematics.
In 1973, the two threads intersected for the first time. Fischer Black, Myron Scholes, and Robert Merton published the options pricing model, and the Chicago Board Options Exchange (CBOE) opened that same year. The Black-Scholes model gave the entire derivatives market an "anchor" -- options were no longer gambling instruments but financial contracts that could be precisely priced with a formula. It directly catalyzed explosive growth in the options market and moved "model-driven trading" from academic papers to trading floors.
2.2 Mathematicians Invade Wall Street (1980s)
If the 1970s were the theoretical foundation period, then the 1980s were when quantitative trading was truly "born." Several pivotal events unfolded in quick succession.
1982: James Simons founded Renaissance Technologies. Simons wasn't a finance guy -- he was a world-class mathematician who had worked on code-breaking at the Institute for Defense Analyses and served as chair of the mathematics department at Stony Brook University. The Chern-Simons theory he co-developed with Shiing-Shen Chern was a landmark in differential geometry. A pure mathematician going into investing seemed absolutely crazy at the time.
Six years later, the landscape took further shape. 1988: Simons established the Medallion Fund, and David Shaw founded D.E. Shaw & Co. Shaw was a computer science professor at Columbia University, specializing in parallel computing. These two firms shared a common trait: they didn't hire financial analysts -- they hired mathematicians, physicists, and computer scientists.
Also in the mid-1980s, Morgan Stanley quantitative trader Nunzio Tartaglia assembled a team composed entirely of physicists and mathematicians. They developed "pairs trading" -- finding two stocks with highly correlated price movements, going long one and short the other when the spread deviated from its historical mean, and waiting for reversion. This was the embryo of statistical arbitrage.
This group shared a common belief: the market is not perfectly efficient; there exist "inefficiencies" that can be captured mathematically. Their working methods were fundamentally different from traditional fund managers -- no reading financial statements, no listening to earnings calls, no attending luncheons. Just data, models, and parameter optimization.
The Medallion Fund's subsequent performance proved the power of this belief: according to publicly reported estimates, from 1988 to 2023, its annualized return exceeded 66% (before fees), with approximately 39.1% after fees (Note: Renaissance has never publicly disclosed complete annual return figures; these numbers primarily come from Gregory Zuckerman's "The Man Who Solved the Market" and court document leaks, and may vary by source). To put this in perspective, the S&P 500's annualized return over the same period was about 10%. Medallion's return curve was nearly vertical.
2.3 The Fall and Rebirth of Geniuses (1990s)
The 1990s were quantitative trading's adolescence -- full of genius, hubris, and painful lessons.
In 1994, a hedge fund called LTCM (Long-Term Capital Management) was founded. Its roster was arguably the most star-studded in financial history: founder John Meriwether was a legendary bond trader from Salomon Brothers, and the advisory team included two Nobel laureates -- Myron Scholes and Robert Merton, the very inventors of the Black-Scholes model. LTCM used complex mathematical models for fixed-income arbitrage, with leverage varying over time -- peaking at around 25-30x (some sources cite over 40x at its highest).
For the first few years, LTCM's performance was staggering. Returns of approximately 43% in 1995 and 41% in 1996, with annualized returns exceeding 40% over the first three years. Wall Street was captivated, and banks competed to lend to them.
Then 1998 hit.
That summer, Russia erupted in financial crisis, defaulting on its government bonds. Global markets plunged into panic, and liquidity evaporated overnight. LTCM's models assumed that asset returns followed a log-normal distribution -- but in the real world, extreme events occur far more frequently than a normal distribution predicts. In just a few months, LTCM lost approximately 4.7 billion to under $1 billion.
The Federal Reserve had to coordinate 14 Wall Street banks to put together a $3.6 billion bailout. Why? Because LTCM's positions were so massive that if it were forced into liquidation, the entire financial system could collapse.
LTCM's spectacular collapse left profound lessons for the quantitative trading industry:
Models have boundaries. No matter how sophisticated a mathematical model, it cannot fully capture human panic and greed. The log-normal distribution assumption works well during "normal times," but when a "black swan" event arrives, it delivers a fatal blow.
Leverage is a double-edged sword. 30x leverage means a mere ~3% adverse move can bankrupt you. In extreme markets, such moves can happen within minutes.
Liquidity is not always there. Models assume you can close positions at any time -- but in a crisis, everyone's liquidity disappears simultaneously.
Interestingly, LTCM's failure didn't halt the development of quantitative trading. On the contrary, it pushed the industry to take risk management more seriously. Jim Simons' Medallion Fund formally adopted its current strategy model in 1998, the very year LTCM collapsed -- learning from LTCM's lessons, it built a rigorous risk control system and has never suffered a major loss since.
2.4 The Golden Age of High-Frequency Trading (2000s)
Entering the 21st century, technological advances pushed quantitative trading into an entirely new phase.
In 2001, U.S. stock markets switched from fractional (hexadecimal) quotes to decimal quotes. This seemingly technical change actually opened the door for high-frequency trading -- the minimum price increment shrank from 1/16 of a dollar to 1 cent, meaning more arbitrage opportunities and more granular pricing strategies.
In 2005, the U.S. SEC passed Regulation NMS (National Market System), promoting competition among exchanges. The market structure of multiple exchanges and "dark pools" coexisting created cross-market arbitrage opportunities for high-frequency traders.
By 2006-2007, high-frequency trading already accounted for a substantial portion of U.S. equity trading volume. These traders didn't care about company fundamentals -- they cared about millisecond-level price differences, the microstructure of order books, and who could get their trading instructions to exchange servers faster.
In August 2007, an event shook the entire quantitative industry -- the "Quant Meltdown." Multiple quantitative funds suddenly suffered unprecedented losses in the same week. The cause was strategy crowding: when too many funds used similar quantitative strategies, once an adverse market shift occurred, all funds hit their stop-losses simultaneously, creating a stampede. August 2007 coincided with the early stages of the subprime crisis (BNP Paribas freezing funds, credit market liquidity tightening), and quantitative funds' factor portfolios were indirectly hit, triggering a chain reaction of collective deleveraging.
The Quant Meltdown was the first large-scale warning of "crowding risk" in quantitative trading. It told the industry: when everyone uses the same model, the model itself becomes the risk.
2.5 The Financial Crisis: Quantitative Trading's "Stress Test" (2008)
The 2008 global financial crisis was the most severe test quantitative trading had ever faced.
When the crisis erupted, a vast number of quantitative models failed simultaneously. They assumed asset returns followed a log-normal distribution, but reality was that on the day Lehman Brothers collapsed, the Dow Jones Industrial Average fell over 500 points in a single day -- an event whose probability under the assumed distribution was roughly "once in the age of the universe."
Quantitative strategies hit stop-losses en masse, causing liquidity to dry up abruptly. Prices didn't decline gradually -- they fell off a cliff. The models said "this is impossible"; the market said "I don't care."
But there was one exception: Jim Simons' Medallion Fund still achieved approximately 80% positive returns in 2008.
Why? Because Medallion's strategy didn't rely on a single model. It had hundreds of mutually independent signal sources, each rigorously statistically tested. More importantly, its risk control system didn't rely on simple distributional assumptions -- it used more conservative tail risk estimates. When other funds were knocked down by "black swans," Medallion actually found more trading opportunities amid the volatility.
The 2008 financial crisis's impact on the quantitative trading industry was profound: it made the industry acutely aware of model risk and drove advances in tail risk management and stress testing. But it also proved that excellent quantitative strategies can not only survive a crisis but profit from it.
2.6 Flash Crashes and the AI Invasion (2010s)
At 2:45 PM on May 6, 2010, the Dow Jones Industrial Average plunged nearly 1,000 points in just minutes -- roughly 9% of its index value evaporated instantly. Some stock prices momentarily fell to one cent or surged to $100,000. Markets returned to normal about 20 minutes later.
This was the "Flash Crash" -- certified by Guinness World Records as the "biggest stock market crash caused by automated trading."
Subsequent investigation revealed the trigger was a large institution dumping a massive quantity of E-Mini S&P 500 futures contracts via algorithm in a single wave. High-frequency traders withdrew liquidity as the market fell, and the chain reaction among algorithms caused liquidity to vanish in an instant. The entire process was "machine vs. machine" -- no human was involved in any decision.
The Flash Crash triggered a global regulatory review of high-frequency trading. The U.S. SEC and CFTC launched a joint investigation, exchanges later introduced individual stock circuit breakers (the LULD rule, formally implemented in 2012), and regulators began requiring algorithmic traders to register and report.
But the more important trend of the 2010s was the deep infiltration of AI and machine learning into quantitative trading.
Traditional quantitative strategies relied on linear models and predefined factors -- value, momentum, size. But from the early 2010s, machine learning methods like support vector machines and random forests were introduced. After the deep learning breakthrough of 2012 (AlexNet), neural network architectures like LSTM and Transformer also began to be applied to stock price prediction.
In the mid-2010s, factor investing became mainstream. Smart Beta ETFs (rules-based funds that systematically capture investment factors like value and momentum) proliferated, and Fama-French's three-factor/five-factor models went from academic papers to blueprints for product design. Quantitative trading expanded from a handful of elite institutions to mutual funds, ETFs, and even individual investors.
In the late 2010s, another important shift was the rise of "alternative data." Satellite imagery could track parking lot vehicle density to predict retail earnings; credit card data could reveal consumption trends in advance; social media sentiment could capture market panic and greed. Quantitative funds were no longer just "reading data" -- they were "hunting for data."
2.7 The Present: The AI Arms Race (2020s)
Quantitative trading in the 2020s can no longer be captured by the word "quantitative" alone -- it's more like a global AI arms race.
The COVID-19 pandemic in 2020 caused violent global market fluctuations, and many quantitative models were once again tested by extreme conditions. But this time, the industry recovered faster. Deep learning, reinforcement learning, and large language models (LLMs) were rapidly introduced into strategy development.
In China, a dramatic story was unfolding. High-Flyer Quantitative (幻方量化) was founded in 2015, built its first AI model in 2016, began large-scale AI computing infrastructure investment in 2017, and constructed the "Firefly I" and "Firefly II" high-performance computing clusters with a large number of NVIDIA A100 chips. Then in November 2023, High-Flyer launched DeepSeek-V1 -- and after multiple iterations, DeepSeek-R1 set the world abuzz in early 2025.
A quantitative hedge fund had built a world-class AI model. This was no coincidence -- because quantitative trading and building large models share highly overlapping foundational capabilities: massive data processing, high-performance computing, pattern recognition, and signal extraction. High-Flyer's founder, Liang Wenfeng, realized early on: the ultimate form of quantitative investing isn't building better quantitative strategies -- it's building stronger AI.
On May 10, 2024, Jim Simons passed away at the age of 86. The story of the "King of Quant" came to a close, but the industry he pioneered is evolving at unprecedented speed.
The current quantitative trading landscape exhibits several notable features:
Pervasive AI integration. From factor mining to signal generation, from execution optimization to risk modeling, AI is replacing traditional statistical methods. Recent arXiv papers show that LLMs are being used to build financial trading agents, with sentiment analysis models achieving 74.4% accuracy on U.S. financial news -- far surpassing the 50.1% of traditional dictionary-based methods.
The alternative data arms race. Satellite imagery, credit card data, supply chain data, weather data, IoT data -- quantitative funds are investing ever more heavily in data sources. But the problem is, when everyone uses the same alternative data, "alternative" becomes "mainstream."
Intensifying strategy crowding. The early 2024 micro-cap stock crash in China's A-share market is a textbook case: massive quantitative capital crowded into micro-cap strategies, and the concentrated sell-off in January 2024 triggered a stampede, with the micro-cap index plummeting for days and quantitative strategies suffering widespread losses. This is not an isolated incident -- globally, factor crowding and strategy homogenization have become the industry's biggest hidden concern.
Tightening regulation. In 2024, the China Securities Regulatory Commission (CSRC) officially issued the Regulations on Programmatic Trading in the Securities Market, implementing full-chain oversight of quantitative trading under a "report first, trade later" framework. Lingjun Investment was restricted from trading for 3 days and faced public censure for selling 2.567 billion RMB worth of Shanghai and Shenzhen stocks in a single minute on February 19, 2024. This event became an iconic turning point in quantitative industry regulation.
2.8 The Unique Trajectory of Chinese Quantitative Trading
China's path of quantitative trading development differs markedly from that of the United States.
2010 was year zero. The CSI 300 stock index futures officially began trading, marking the establishment of China's quantitative trading infrastructure. Early quantitative strategies were simple -- primarily cash-futures arbitrage.
2012-2015 was the startup phase. A batch of returnees with experience at overseas quantitative firms came back to China to start businesses: Ubiquant (九坤投资, founded 2012), MingHong (明汯投资, 2014), Lingjun Investment (灵均投资, 2014), and High-Flyer Quantitative (幻方量化, 2015) were established in succession.
2015 was a critical juncture. China's stock market experienced a crash from June to August, and quantitative trading was partially blamed. The China Financial Futures Exchange imposed extremely strict restrictions on stock index futures -- raising margin requirements to 40% and limiting intraday positions to 10 contracts. This was devastating for quantitative hedging strategies, forcing the industry to pivot toward equity multi-factor and CTA strategies.
2019-2023 was the explosive growth period. Stock index futures restrictions were gradually relaxed, and the number of quantitative hedge funds managing over 10 billion RMB grew from single digits to 30-40. Quantitative trading's share of A-share daily trading volume rose from under 5% to 10%-20%. The "Big Four" structure emerged -- High-Flyer, Ubiquant, MingHong, and Lingjun.
2024 was the turning point. A regulatory storm arrived in full force. The new "Nine Guidelines" explicitly called for stronger oversight of high-frequency quantitative trading, and the Regulations on Programmatic Trading in the Securities Market were formally implemented. Lingjun Investment's "market crash" incident became the industry's watershed moment.
What makes Chinese quantitative trading unique is that it grew up in a market dominated by retail investors. Retail investors account for the overwhelming majority of A-share market participants by number, and while institutional participation is steadily rising, retail trading still constitutes a significant proportion. The information and technology asymmetry between quantitative trading and retail investors is one of the root causes of tightening regulation.
III. Horizontal Analysis: The Competitive Landscape
3.1 Global Landscape: Who Rules the Game?
The global quantitative trading landscape can be summarized in one word: highly concentrated.
As of 2024, total global hedge fund assets were approximately 770 billion, representing approximately 50% of total quantitative fund assets. This is a winner-take-all market.
Renaissance Technologies is the legend among legends. The Medallion Fund's long-term performance (detailed in the vertical analysis) makes it the most profitable hedge fund in history. Its success formula is simple: hire top scientists rather than finance professionals, invest heavily in data infrastructure, maintain a strict culture of secrecy, and continuously iterate models. But Medallion is only open to internal employees -- outside investors cannot participate. It doesn't need your money.
Two Sigma is the exemplar of the "tech company" hedge fund model. About two-thirds of its employees are engineers and scientists. It has developed proprietary data platforms and backtesting frameworks and actively uses alternative data. However, around 2024-2025, the two founders entered arbitration over disagreements about the company's direction, causing management turbulence.
Citadel is the model of a dual-engine approach: market-making plus hedge fund. Ken Griffin founded it in 1990; Citadel Securities handles approximately 25% of U.S. equity trading volume, making it one of the world's largest market makers. In 2024, its flagship Wellington fund returned 15.1%, and since inception it has generated $74 billion in cumulative returns for investors.
D.E. Shaw is known as "intellectually intensive." Founder David Shaw is a computer science professor at Columbia University, and Jeff Bezos (Amazon's founder) once worked there early in his career. In 2024, its flagship Composite fund returned 18%, and its Oculus macro fund returned 36%, marking its best annual performance since inception.
Millennium Management is the benchmark for the multi-strategy platform model (Pod model). It recruits a large number of independent investment teams, each managing autonomously, with the platform providing infrastructure and risk control. In 2024, it returned 15%, with assets doubling over the past five years.
Bridgewater Associates is one of the world's largest hedge funds (AUM approximately $90-125 billion, fluctuating in recent years) and a pioneer of macro quantitative investing. Ray Dalio's "Principles" culture has profoundly shaped the firm's operations. In 2024, Bridgewater's China fund returned 35%, standing out in performance.
Man Group / AHL is the world's largest publicly traded hedge fund company (AUM approximately $175.7 billion) and a pioneer of CTA strategies. AHL has been systematically trading since 1987, and its trend-following strategies perform strongly in trending markets.
Jump Trading, Tower Research, Hudson River Trading are the three giants of high-frequency trading. Jump Trading has about 100 employees, covers multiple asset classes, and uses FPGA hardware acceleration to achieve nanosecond-level latency. Tower Research was fined $67.5 million by the CFTC in 2020 for "spoofing." HRT's inaugural intern class produced Scale AI founder Alexandr Wang.
3.2 China's Landscape: Who's Leading?
China's quantitative hedge fund landscape can be summarized as "Big Four + Chasers."
High-Flyer Quantitative is the largest quantitative hedge fund (60-80 billion RMB), registered in Ningbo and operating in Hangzhou. But its biggest impact isn't in quantitative trading -- it's in AI. High-Flyer's DeepSeek large model attracted global attention in early 2025, demonstrating that the technical capabilities built up by quantitative hedge funds in AI R&D can break through boundaries.
Ubiquant is one of the earliest teams engaged in quantitative trading in China (founded 2012). Founder Wang Chen graduated from Tsinghua University's computer science department. In 2025, in collaboration with Microsoft, Ubiquant successfully reproduced DeepSeek-R1, discovering that language mixing significantly degrades reasoning capabilities.
MingHong founder Qiu Huiming has over 20 years of investment experience and previously served as a fund manager at the globally renowned hedge fund Millennium. MingHong's distinguishing features are its overseas hedge fund experience and strong team research capabilities.
YanFu Investment is a rising star (founded 2019), established by Gao Kang, who has an overseas quantitative background. Its products have repeatedly reached new highs, with the "YanFu Water Drop Neutral No. 1" and other products hitting record net values in September 2024.
Lingjun Investment was originally one of the "Big Four," but on February 19, 2024, it sold 2.567 billion RMB worth of Shanghai and Shenzhen stocks in a single minute, resulting in a 3-day trading restriction and a public censure proceeding by the exchange. This incident directly drove regulators to scrutinize quantitative trading more strictly.
Chengqi Asset Management (Shenzhen) and QiLin Investment are representatives of the second tier, with management scales in the 20-50 billion RMB range.
3.3 Strategy Types: Different Weapons, Different Battlefields
Quantitative trading strategies can be categorized by trading frequency into three major types:
High-Frequency Trading (HFT) is "millisecond warfare." Holding periods range from milliseconds to seconds, annualized turnover is extremely high, and capacity is very small (tens of billions of dollars). Representatives include Jump Trading, Tower Research, and HRT. Advantages: low risk, stable returns, market-direction-neutral. Disadvantages: extremely high technical barriers, cutthroat competition, continuously compressing profit margins. Requires FPGA hardware acceleration, co-location, kernel bypass, and other cutting-edge technologies.
Mid-frequency strategies are "intraday games." Holding periods range from hours to days. Representatives include Renaissance's Medallion Fund and Two Sigma. Advantages: excellent risk-adjusted returns, suitable for AI/ML applications. Disadvantages: requires powerful computing and data, strategy crowding.
Low-frequency strategies are "macro bets." Holding periods range from weeks to months, with enormous capacity (hundreds to thousands of billions of dollars). Representatives include Bridgewater, AQR, and Man Group. Advantages: large capacity, suitable for big capital. Disadvantages: relatively lower returns, heavily influenced by macroeconomic factors.
Specific strategy types include:
Statistical arbitrage / market neutrality exploits statistical relationships between assets for hedged trades, independent of market direction. Advantages: stable returns. Disadvantages: strategy crowding leading to alpha decay, correlations may break down in extreme conditions. Renaissance's Medallion fund and many Chinese quantitative hedge funds employ this strategy.
CTA / managed futures trades in futures markets based on trend following. Advantages: low correlation with traditional assets, crisis alpha (often performs well when markets crash). Disadvantages: losses when trends are unclear. Man Group/AHL is the benchmark for CTA strategies.
Factor investing constructs portfolios based on factors like value, momentum, quality, and volatility. Strong theoretical foundation (Fama-French models), high interpretability; but factor crowding and factor timing rotation are major risks. AQR Capital is the representative of factor investing.
Machine learning / deep learning strategies use ML/DL models to mine non-linear alpha signals. Can discover patterns that traditional methods miss, but overfitting risk is high and interpretability is poor. Two Sigma and High-Flyer are representatives.
Alternative data strategies utilize non-traditional data sources like satellite imagery, credit card data, and social media. Unique information advantage, but high data acquisition costs and compliance risks. Two Sigma, Point72, and WorldQuant are representatives.
3.4 Technology Stack: The Hardware of the Arms Race
The quantitative trading technology stack can be divided into five layers:
Data layer: Market data (Level 1/2/3 tick data) is the foundation; fundamental data is standard for mid-frequency strategies; alternative data (satellite imagery, social media, supply chain data) is the frontier. The Chinese market also has unique data sources such as the Dragon and Tiger List, margin trading data, and northbound capital flows.
Compute layer: HFT uses FPGA for nanosecond-level processing; mid-frequency strategies use GPU clusters (A100/H100) for model training; low-frequency strategies can use ordinary servers. High-Flyer built its own "Firefly I" and "Firefly II" supercomputing clusters.
Execution layer: Co-location is standard for HFT; kernel bypass achieves microsecond-level latency; GPU acceleration is used for model inference.
Risk control layer: From the strategy level (individual strategy stop-losses) to the portfolio level (cross-strategy risk aggregation) to the execution level (order checks) to the system level (failover) to the compliance level (regulatory reporting) -- a five-layer risk control system where every layer is indispensable.
Backtesting layer: Large institutions build their own backtesting systems; small and medium institutions use open-source frameworks (Zipline, Backtrader, vnpy); brokerages provide trading platforms like QMT.
3.5 Market Size: Let the Numbers Talk
Global: Total hedge fund assets approximately 1.35-1.8 trillion. The algorithmic trading software market was 6.22 billion by 2032 (CAGR 9.4%). Quantitative/algorithmic trading accounts for approximately 60-70% of U.S. equity trading volume.
China: Quantitative hedge fund AUM approximately 1.5-1.8 trillion RMB (2024), with 30-40 funds managing over 10 billion RMB. Quantitative trading accounts for approximately 10-20% of A-share trading volume. There are over 1,000 quantitative investment firms in China.
China vs. the World: U.S. quantitative trading accounts for 60-70%, while China is only 10-20%, leaving enormous room for growth. However, the largest Chinese quantitative hedge fund tops out at about 80 billion RMB (~$11 billion), far below the trillion-dollar level of global top firms. Chinese strategies are primarily based on volume-price and index enhancement, whereas global top firms employ highly diversified strategies.
IV. Horizontal-Vertical Intersection Insights
4.1 How History Shaped Today's Competitive Positions
Renaissance Technologies' legendary status in the industry traces back to a key decision Jim Simons made in 1988: hire scientists, not finance people. This decision seemed heretical at the time, but it built Renaissance's deepest moat -- a talent barrier. While other funds recruited MBAs from business schools, Renaissance recruited PhDs from math and physics departments. These people's thinking was fundamentally different from traditional finance professionals -- they didn't care about "stories," only about "signals."
D.E. Shaw's "intellectually intensive" positioning also has historical roots. David Shaw is a computer science professor who brought computer science thinking -- systems design, algorithm optimization, parallel computing -- into investment management. Two Sigma's co-founder John Overdeck came from D.E. Shaw, inheriting this "technology-driven" gene and pushing it to the extreme -- two-thirds of employees are engineers and scientists.
Citadel's dual-engine model (hedge fund + market-making) stems from an insight Ken Griffin had in 1990: market-making can generate stable cash flow, providing low-cost funding for the hedge fund business. This model gives Citadel greater resilience during market volatility.
4.2 Longitudinal Comparison of Competitors: Different Origins, Different Fates
Viewing the world's major quantitative institutions on a timeline, their origins determined their present forms.
Renaissance and D.E. Shaw were both born in the 1980s, both centered on "scientist-driven" principles. But Renaissance chose a "closed ecosystem" -- Medallion is only open to internal employees, with strategies kept highly confidential. D.E. Shaw chose an "open ecosystem" -- cultivating a vast amount of talent (Two Sigma founders, Amazon's Bezos both worked there), but also to some extent "leaking" its methodology.
Two Sigma was born in 2001, nearly 20 years after Renaissance. Its advantage is standing on the shoulders of giants -- inheriting D.E. Shaw's technology gene while leveraging the explosion of internet and big data technologies in the 2000s. But its disadvantage is lacking Renaissance's "zero-to-one" originality -- it's more about optimizing existing paradigms than creating new ones.
Citadel's Ken Griffin came from a trading background, not science. His DNA is "making money," not "exploration." This determined that Citadel's strategies are more pragmatic and diversified, but also means it's unlikely to achieve perfection in any single area. Citadel Securities' market-making success partially compensates for this "not quite extreme enough" shortcoming.
Chinese quantitative hedge funds have taken a completely different path. High-Flyer's founder Liang Wenfeng is neither a mathematician nor a finance person -- he's more of a tech geek. He realized early on: the ultimate form of quantitative investing isn't building better quantitative strategies -- it's building stronger AI. This understanding led High-Flyer to gradually transform from a quantitative hedge fund into an AI R&D institution, and DeepSeek's birth is the product of this vision.
Ubiquant's Wang Chen graduated from Tsinghua's computer science department; his DNA is "engineering," not "science." Ubiquant's hallmark is solid engineering capability and continuous technical iteration, rather than disruptive theoretical innovation.
MingHong's Qiu Huiming has Millennium work experience; his DNA is "hedge fund operations," not "technological breakthroughs." MingHong's strengths are team research capability and the localization of overseas experience.
4.3 The Historical Roots of Advantages
Renaissance's core advantage -- its "black box strategy" -- traces back to Jim Simons' code-breaking background. The core idea of cryptography is: extracting signals from noise. This is precisely the essence of quantitative trading. Simons brought this thinking to finance, establishing Renaissance's "pattern recognition" methodology.
High-Flyer's AI advantage traces back to a 2016 decision: use deep learning for stock trading. While most Chinese quantitative hedge funds were still using traditional multi-factor models, High-Flyer was already all-in on AI. This "two years ahead" time advantage is decisive in the AI domain.
Citadel's market-making advantage traces back to the founding of Citadel Securities in 2002. Ken Griffin realized early that the technology infrastructure for market-making could be shared with the hedge fund business, creating synergies. This "dual-engine" model gave Citadel a unique competitive edge.
4.4 The Historical Roots of Disadvantages
Renaissance's "closedness" is a double-edged sword. Medallion is only open to internal employees, and outside investors cannot participate -- this has allowed Renaissance to earn the most money, but has also limited its scale growth. Other funds (RIEF, RIDA, etc.) targeting external investors have performed far worse than Medallion.
Two Sigma's "openness" also has its costs. The two founders entered arbitration around 2024-2025 over disagreements about the company's direction, exposing a common governance problem in "technology-driven" companies: when two tech geniuses disagree on direction, there's no "final arbiter."
Lingjun Investment's "market crash" incident has its roots in its "mid-to-high frequency strategy" DNA. Mid-to-high frequency strategies pursue speed and scale -- completing the maximum trading volume in the shortest time. This DNA is an advantage in normal markets but becomes a fatal weakness in extreme markets.
4.5 Future Scenarios: Three Scripts
Most Likely Scenario: AI Deepening and Regulatory Balance
Quantitative trading will continue to evolve toward AI and intelligence. LLMs will be more widely applied to strategy research and trading decisions, and alternative data usage will become more widespread. Meanwhile, global regulation will continue to tighten, but won't be a blanket ban -- the goal of regulation is "regulated development," not "eliminating" quantitative trading. Chinese quantitative hedge fund AUM will double in the next 3-5 years, and quantitative trading's share of A-share trading volume will rise from 10-20% to 30-40%.
Logic: Exponential progress in AI technology (computing power growth, model capability improvement), continued opening of China's capital markets (more financial instruments, more data sources), and gradual improvement of the regulatory framework (giving the industry clear rules to follow).
Most Dangerous Scenario: AI Convergence and Systemic Risk
As more quantitative funds use similar AI models and data sources, the risk of strategy homogenization will continue to increase. In some extreme market event, all AI strategies might simultaneously trigger sell signals, causing a stampede larger than the 2024 Chinese micro-cap crash. If such a stampede occurred across global markets -- simultaneously triggering quantitative stop-losses in the U.S., Europe, and Asia -- it could spark a systemic financial crisis.
Logic: Lessons from the 2024 Chinese micro-cap crash, the precedent of the 2010 Flash Crash, difficulty in assessing risk due to the "black box" nature of AI models, and increasing systemic linkages from the growing share of global quantitative trading.
Most Optimistic Scenario: Quantitative Trading Empowering the Real Economy
Quantitative trading's AI capabilities will be applied to broader financial domains -- risk management, asset pricing, and inclusive finance. AI technologies incubated by quantitative hedge funds (such as High-Flyer's DeepSeek) will feed back into the real economy, driving AI applications in healthcare, education, manufacturing, and other fields. Quantitative trading will transform from a "zero-sum game" to "value creation."
Logic: DeepSeek's success proves that quantitative hedge funds' AI capabilities can cross boundaries, the rapid development of the global AI industry, and financial regulation guiding capital from "virtual" to "real."
Data Reliability Note
High reliability (primary sources)
- 2010 Flash Crash investigation findings: from SEC and CFTC joint investigation report, Guinness World Record certification
- China Securities Regulatory Commission's "Securities Market Programmatic Trading Management Regulations": from CSRC official announcement
- Lingjun Investment disciplinary action: from exchange public censure and East Money reporting
- SEC Reg NMS, Reg SCI regulations: from SEC official website
- LTCM collapse: from extensive financial history literature and Federal Reserve rescue records
- 2024 hedge fund performance data: from Sina Finance (QIML reporting) compilation
Medium reliability (secondary sources but cross-verifiable)
- Medallion Fund returns (66% pre-fee / 39.1% post-fee): from Gregory Zuckerman's "The Man Who Solved the Market" and court document leaks; Renaissance has never publicly confirmed
- LTCM leverage (25-30x): from multiple financial history works; specific figures varied by period
- Global hedge fund AUM and quant strategy share: from industry reports and media compilations; figures vary slightly by source
- China quant hedge fund AUM (RMB 1.5-1.8 trillion): from Simuwang.com and industry research reports
- Individual firm AUM data: from Tencent News, Huxiu, and other media compilations; specific figures may differ by reporting date
Lower reliability (cite with caution)
- Bridgewater AUM (~$90-125 billion): fluctuates in recent years; specific figures depend on timing and source
- Two Sigma founders' arbitration: from Huxiu reporting; details unverified
- Quant trading share of A-share volume (10-20%): estimates vary widely across sources
- Sentiment analysis model accuracy (74.4% vs 50.1%): from arXiv preprint, not peer-reviewed
V. Sources
Vertical Analysis Sources
Horizontal Analysis Sources
| # | Source | URL |
|---|---|---|
| 14 | Sina Finance (QIML): 2024 Global Hedge Fund Performance | https://finance.sina.com.cn/money/fund/jjzl/2025-01-06/doc-ineczakc2666144.shtml |
| 15 | Ofweek: 2024 Quantitative Investment Industry Report | https://mp.ofweek.com/finance/a656714951537 |
| 16 | Simuwang: China Quantitative Hedge Fund Rankings | https://simuwang.com |
| 17 | Tencent News: 2024 Global Quantitative Hedge Fund AUM Rankings | https://news.qq.com/rain/a/20240531A04OTT00 |
| 18 | Credence Research: Algorithmic Trading Software Market | https://credenceresearch.com/report/algorithmic-trading-software-market |
| 19 | Huxiu: Two Sigma Management Changes | https://huxiu.com/moment/1029333.html |
| 20 | Eastmoney: Lingjun Investment Incident | https://finance.eastmoney.com/a/202402202989619652.html |
| 21 | Sohu: High-Flyer / DeepSeek | https://sohu.com/a/848281393_122066678 |
| 22 | iFeng: Ubiquant Reproduces DeepSeek | https://tech.ifeng.com/c/8hSN3gMHyen |
Supplementary Sources
Report completed: May 8, 2026