The Eight Schools of Quant Trading

Quant trading is not a monolithic thing. Citadel Securities processes tens of millions of orders per second; Renaissance Medallion compounded at 39% net for 30 years; AQR manages $140B in factor strategies — they all call themselves “quants” but they’re not really doing the same thing. This guide breaks the field down by school and lays out, for each, the core idea, representative firms, market share, holding period, hardware / data / talent intensity, and the strengths and limits — so you can figure out “which kind of quant do I actually want to do” before stepping in.

Framework — four dimensions of quant

“Quantitative trading” encompasses every capital flow guided by mathematical models, from sub-microsecond market making to annual-rebalance long-term factors. Different schools look like “code + data,” but they differ enormously across four dimensions — getting these four straight before diving into any single school keeps you from getting confused by the catch-all term.

1. Holding period. From microseconds (HFT), minutes-to-hours (StatArb), days-to-weeks (event-driven / intraday factors), months-to-quarters (CTA / macro), all the way to years (Factor / Risk Premia). Period determines source of return, capacity, and data / hardware requirements. Shorter → more dependent on infrastructure; longer → more dependent on research depth.
2. Source of return. There are really only four: liquidity provision (market making), mispricing (arbitrage), risk bearing (risk premia), forecasting alpha. Each school mixes them in different proportions. “Whose money am I actually winning” is the foundational question.
3. Capacity. HFT ≤ $10B, StatArb ~$50-100B, single-strategy CTA ~$20B, factor strategies $1T+. Lower-frequency strategies have higher capacity — the structural reason HFT lives in proprietary trading while low-frequency lives in asset management. Medallion is hard-capped at ~$10B = a ceiling for HFT-class capacity.
4. Infrastructure intensity. A continuum from a single laptop (factor) to transcontinental microwave + FPGA (HFT). HFT’s technical bar is like building rockets; factor’s is like writing a book. Misjudging this forces small teams onto tracks they fundamentally can’t compete on. See the “Hardware / Data / Talent” section.

Bottom Line · Pick your school first, then talk strategy.

“Doing quant” is a category as broad as “doing internet.” A two-person team with $1M can’t outbid Citadel for market share (the hardware delta is 5 orders of magnitude); but it can absolutely run CTA trend-following or multi-factor strategies. Choosing the right school is the most important first step on the way in.

High-Frequency Market Making — the microsecond arms race

HFT (High-Frequency Trading) and market making are strictly distinct in the academic literature, but in practice the top firms do both — quoting two-sided markets to earn the spread + sniping others’ quotes for arbitrage. This is the “hardest” school of quant: hard hardware, hard people, hard profits.

Core idea

Markets naturally have a Bid-Ask spread — some people want to buy immediately, some want to sell immediately. Market makers quote two-sided prices to provide instant liquidity and collect this spread as a “service fee.” A few cents per trade × hundreds of millions of trades per day = a steady cash flow.

The HFT extensions: Latency Arbitrage — when the same stock is briefly mispriced across NYSE and Nasdaq for milliseconds, the faster machine eats the slower one’s quote; Order Book Prediction — using order placement / cancellation microstructure to predict the next 100ms of price action.

Representative firms

• Citadel Securities — handles ~40% of US retail order flow, global #1 in ADV
• Virtu Financial — publicly listed market maker; 1,500+ consecutive winning trading days on record
• Jump Trading — early microwave network pioneer · Chicago-NY private line
• Jane Street — ETF market-making leader, OCaml-flavored, top of the campus-recruiting heap
• Hudson River Trading (HRT) · DRW · IMC · Tower Research · Two Sigma Securities

Market share and scale

Typical P&L

A common stat for market makers: Sharpe ratio of 8-12 (annualized return / annualized vol), profitable nearly every day, with maybe 1-2 losing days per month. But capacity is extremely limited — return drops sharply once capital grows past a certain point. That’s the core reason HFT stays in proprietary trading rather than asset management.

Pros and cons at a glance

1. Strengths. Extremely high Sharpe (8-12+) · low/zero correlation to traditional assets · benefits from crisis years (liquidity demand spikes) · structurally hard to poach (code + hardware + people are a systemic moat).
2. Limits. Hard capacity ceiling · regulatory risk (SEC/FINRA fines, market manipulation charges) · enormous arms race spending ($100M+ annual hardware capex) · poor optics (“extracting capillary spreads”).

⚠ Common newcomer misconception · HFT is not “writing a strategy + adding leverage.”

A retail trader self-teaching Python and renting a VPS isn’t doing HFT. Real HFT requires: microwave / laser private lines, PCIe-direct exchange NICs, custom FPGA hardware, hand-tuned assembly, co-located cabinets, and exchange membership — fall short on any one of these by an order of magnitude and you’re outclassed instantly. Retail should not attempt HFT.

Statistical Arbitrage — the school Medallion made famous

Statistical Arbitrage (StatArb) is the most “classical” of the quant schools. Core idea: use statistical models to find temporary mispricings and bet on mean reversion. From its 1980s origins under Nunzio Tartaglia’s APT group at Morgan Stanley, through Renaissance / DE Shaw / Two Sigma scaling to hundreds of billions.

Core idea

The earliest version is called pairs trading: when two related stocks (e.g. Coca-Cola / Pepsi) diverge from their historical mean, you go long the cheap one and short the expensive one, betting on convergence. This was later generalized to multi-stock, multi-factor models: decompose stocks into dozens or hundreds of factor exposures, and arbitrage each dimension using historical means + volatility.

Typical modern StatArb holding periods: minutes to days. Medallion is reputedly mostly intraday; DE Shaw / Two Sigma cover minutes to weeks; some mid-frequency funds operate on day-to-week scales.

Representative firms

• Renaissance Technologies · Medallion — since 1988, 30 years of 39% net annualized, internal only, capped at ~$10B. The spiritual icon of every quant fund.
• DE Shaw — since 1988, ~$60B AUM, full product line. David Shaw himself is a computer scientist.
• Two Sigma — since 2001, ~$65B AUM, early ML adopter.
• Citadel (Wellington / Tactical) — multi-strategy flagship; the quant arm is a major driver.
• PDT Partners — descends from the Morgan Stanley APT lineage, founded by Peter Muller.
• WorldQuant — Igor Tulchinsky (former Millennium quant head), the “crowdsourced alpha” model.

Scale and standing

P&L profile

The standard StatArb picture: market-neutral + highly diversified (hundreds to thousands of stocks). Per-trade hit rate may only be 51-53%, but with massive trade count the Central Limit Theorem pulls the overall Sharpe up to 3+.

The trouble is “pattern collapse” — the 2007-08 “Quant Quake,” where StatArb funds had crowded into similar factors, all deleveraged at once and trampled each other → Goldman GEO fell 30% in a week. This is the crowding risk built into StatArb.

Pros and cons at a glance

1. Strengths. Sharpe 2-4 · market neutral · capacity to $50B+ · academically transparent methodology (mean reversion + cointegration) · strong compounding effect on data / research.
2. Limits. Crowding risk · factor decay (alpha half-life 1-3 years) · high infrastructure cost · top talent is gated behind “Renaissance ivory tower”-style closed access.

CTA Trend Following — the fifty-year-old school

CTA (Commodity Trading Advisor) is the oldest systematic school, registered under CFTC supervision in the US. The “Turtle Traders,” Bill Dunn, and John Henry started running these strategies from the 1970s-80s. The core idea is simple enough to write on a single sheet: once a price trend forms it persists, so follow it + cut losses strictly.

Core idea

Put 50-150 global futures (equity indices, rates, FX, energy, metals, ags) into a basket; for each, use moving averages, breakouts, momentum rules to determine direction and sizing. When a trend holds, sit on it; when reverse signals arrive, flip or flatten.

The essence is a convex strategy — frequent small losses, rare large gains. Most years are slightly positive or slightly negative, with outsized returns in crisis years (2008 AHL +40%, 2022 CTA index +21%).

Representative firms

• Man AHL — London origin; AHL Diversified is a benchmark CTA strategy.
• Winton — founded by David Harding (one of AHL’s early co-founders before leaving).
• Aspect Capital — Anthony Todd (also from AHL).
• DUNN Capital — Bill Dunn since 1974, five decades across generations.
• Graham Global · Millburn · Chesapeake · Transtrend · Lynx Asset Management
• AQR Managed Futures — academic-flavored CTA, a $10B+ “mutual fund version.”

Scale and performance

P&L profile

CTAs are characterized by positive skew — hit rates of just 35-45%, but the average size of winning trades far exceeds that of losers (3-5:1). “Let your winners run, cut your losers” — that old-school maxim is the CTA in a sentence.

The biggest risk is prolonged drawdown: CTAs as a group underperformed bonds for nearly 6 years from 2013-2019, a “trend drought” that shuttered many funds. They revived from 2020-2022, but it’s still a “wait for the wind” strategy, requiring patient clients.

Pros and cons at a glance

1. Strengths. Clear crisis alpha · low correlation to stocks and bonds · transparent and explainable · large capacity ($50B+ per strategy) · low hardware requirements.
2. Limits. Low Sharpe (0.5-0.8) · grinding long drawdowns (6-year plateaus are not unusual) · whipsaws in choppy markets · entirely dependent on the existence of trends.

Systematic Macro — coding Dalio’s logic

Systematic macro sits between CTA and StatArb — slow trades on macroeconomic variables based on causal economic reasoning, but executed by code rather than by a portfolio manager’s discretion. The flagships are the quant portion of Bridgewater Pure Alpha, AQR Macro, and Graham Global.

1. Core idea. Encode causal macro relationships like “Fed hikes → USD up → EM FX down” as rules, calibrate parameters from historical data, and rebalance periodically. The difference from CTA: CTAs look at price, macro looks at fundamentals. The difference from discretionary macro: no PM discretion — fully systematic.
2. Representative firms. Bridgewater Pure Alpha (quant portion), AQR Macro, Graham Global, AHL Evolution, Man Numeric macro products.
3. Scale. Global systematic macro AUM ~ $150B. ~30-40% of Bridgewater’s $125B is systematic.
4. Holding period. Weeks to quarters. Portfolio turnover is much lower than CTA; parameter adjustments feel more like “macro canvas” than “signal.”
5. Return profile. Sharpe 0.7-1.2; Pure Alpha long-term net annualized ~10-12% (below Medallion but 10× the capacity).
6. Biggest challenge. “Macro causality” breaks down during QE / zero rates / policy abnormalities. In 2022-2023, the Fed’s aggressive hiking inverted historical “rates vs. FX” correlations and caught many macro models flat-footed.

Factor / Risk Premia — the most mainstream, most academic, cheapest

Factor investing is the school where academia and practice are most tightly integrated. After Fama-French published the three-factor and five-factor models, Value / Momentum / Size / Quality / Low-Vol and related factors were industrialized into funds, ETFs, and smart-beta products — the only quant school that has trickled down into ordinary investors’ pockets.

Core idea

Rank stocks (or other assets) by a fundamental or price characteristic, go long the high-ranked and short the low-ranked (long-short), or just long the high-ranked (long-only smart beta). Over the long run these factors carry a “risk premium” — excess return earned for bearing a particular kind of risk.

The six classic factors:

• Value: long low P/B, short high P/B
• Momentum: long past-12-month winners, short losers
• Quality: high ROE, low leverage, stable earnings
• Size: small cap > large cap (historical premium; broke in the 2010s)
• Low-Vol: low-volatility stocks haven’t underperformed high-vol over the long run
• Carry: arbitrage of interest rate differentials / forward basis (cross-asset)

Representative firms

• AQR Capital — founded by Cliff Asness (a Fama student), $140B, the largest factor fund.
• Dimensional Fund Advisors (DFA) — deeply partnered with Fama / French, $775B, built on factor-passive products.
• Research Affiliates — Rob Arnott’s “fundamental indexing” school.
• BlackRock Systematic / SAE · State Street Advanced Research — the quant arms of the megacap asset managers.
• Vanguard / iShares Factor ETFs — MTUM, VLUE, QUAL, USMV are retail-accessible factor ETFs.

Scale · performance

Defining characteristics

Enormous capacity + ultra-low fees. Smart beta ETFs charge 0.15-0.30%, a world apart from hedge fund 2/20. That’s the only reason factor strategies could “trickle down to retail.”

But extended underperformance is normal — Value lagged growth by 30% from 2010-2020, with investors nearly abandoning the academic faith. Then it rallied hard from 2021-2023. Factor investing is a “tax on patience.”

Pros and cons at a glance

1. Strengths. Unlimited capacity · ultra-low fees · academically transparent · accessible via ETF · long-term Sharpe ~0.5 has marginal value vs. stocks and bonds.
2. Limits. 5-10 year underperformance is common · factor crowding · client withdrawal pressure · repeatedly “slapped” in years of extreme AI / tech outperformance.

Machine Learning — a decade of neural nets in finance

Strictly speaking, ML quant isn’t a standalone school — it’s a cross-cutting tool any school can use. But “ML / deep-learning-as-the-core-method” funds have grown into a school of their own, with distinct datasets, infrastructure, and talent profiles. It took off in the late 2010s and after 2020 became standard for every new quant fund.

Core idea

Traditional factor models assume linear relationships (y = a + bx + ε). ML allows non-linear + high-dimensional interactions — Gradient Boosting, XGBoost, Random Forests, LSTMs, Transformers have all been tested on quant.

Typical applications:

• Feed 1,000+ features (financials, sentiment, alt data, price/volume) into a boosting model
• Use NLP (BERT/LLM) to turn news / earnings calls / Reddit posts into features
• Use CV to count cars in satellite images of Walmart parking lots, or shadows on oil tanks
• Reinforcement learning (RL) for optimal execution (TWAP/VWAP optimization, liquidation)

Representative firms

• Renaissance (rumored to have used ML deeply for a long time internally)
• Two Sigma — the most famous ML shop in the industry, with research platforms Halite / Beta open to the public
• WorldQuant — the “Alpha Factory” running tens of thousands of factor searches in parallel
• Voloridge Investment Management — founded by David Vogel, pure ML
• Numerai — crowdsourced data science, “crypto + quant”
• QuantConnect / Quantopian (now closed) — community ML strategy platforms

Reality check

Not as magical as marketed. The real situation:

• Feature engineering > algorithm choice — XGBoost and LSTM differ by < 1% in production
• Sample size is the bottleneck — 30 years of daily data × 5,000 stocks = under 40M samples, a tiny dataset for deep learning
• Backtest washing (overfitting) is rampant — every “neural net Sharpe 3” backtest you’ve heard of has a “live Sharpe 0.5” sequel
• Poor interpretability — risk teams can’t explain “why are we long Kweichow Moutai right now,” which is a real headache for compliance and risk

Infrastructure

⚠ The trap newcomers fall into · “Picking stocks with GPT” / “LLM stock picking” is 99% a scam.

The gap from academic research to production deployment is enormous. A backtest Sharpe of 3.0 turns into a live Sharpe of −0.5 about 70% of the time — for reasons including look-ahead bias, survivorship bias, underestimated transaction costs, and out-of-sample failure. Any “AI quant product” that can’t show three or more years of live trading should be treated as having no track record.

Event-Driven Quant — short-term alpha around catalysts

Event-driven quant sits between StatArb and fundamental quant. Core logic: find a clear catalyst (earnings, M&A, index inclusion, analyst rating change) and systematically bet around the event over a window of days to weeks.

1. PEAD · Post-Earnings Announcement Drift. After a large earnings beat, prices continue drifting in the same direction for 1-2 months. The longest-confirmed academic anomaly, but alpha has decayed from ~1%/month in the 1980s to ~0.2%/month today.
2. Merger Arb. For cash deals: buy the target after announcement, lock in “deal price − current price.” Spreads are typically 1-3%; deal breaks can be −20%+. Win rate 90%+, but extreme tail risk.
3. Index Rebal. In the 4-6 weeks between an S&P / Russell rebalance announcement and effective date, additions rise an average of 5-8%, deletions fall 3-5%. Alpha has narrowed since 2020 but remains stable.
4. Analyst Rev · analyst rating changes. Upgrades deliver ~0.5-1% alpha over 1-5 days; downgrades are stronger. Combined with Institutional Broker’s Estimate System (IBES) data.
5. IPO / Spin-off. Spin-off children outperform over the long run (Joel Greenblatt’s classic strategy). IPOs go the other way — long-term market underperformers.
6. Buyback announcements. Large buyback announcements deliver 3-5% market outperformance over 6-12 months. Filter by buyback size / market cap.

Representative firms

Millennium (some pods of the multi-strat), Point72 (Cubist quant arm), Balyasny, Schonfeld and other “pod shops” use event-driven quant heavily. Standalone pure event-driven funds are rare; most live inside multi-strat platforms.

Crypto Quant — the 24/7 new market

Crypto quant is the only “brand new” market created in the last decade. 2017-2021 was the wild west period when cross-exchange arbitrage could deliver Sharpe 5+ easily; post-2022, traditional HFT firms like Jump, Jane Street, and DRW have moved in and it’s now “cleaner” — but some structural alpha remains.

1. CEX Arb · centralized exchange arbitrage. Across dozens of exchanges (Binance / Coinbase / Bybit / OKX), BTC / ETH price gaps occasionally hit 10-50bps. Pre-2021 it reached 1-5%. Requires capital pre-positioned at each venue + low-latency networking.
2. Funding Rate arbitrage. Perpetual contracts settle funding every 8 hours. Long spot + short perpetual locks in the funding rate. 20-40% annualized during the 2021 bull market; now 2-8%.
3. DEX MEV · Maximum Extractable Value on-chain. Sandwich attacks, arbitrage, liquidations. Since Ethereum’s PoS transition, divided between Builders and Searchers. MEV-boost has extracted $1B+ cumulatively; top searchers make tens of millions a year.
4. Basis Trade. Spot / futures basis arbitrage. BTC’s annualized basis can hit 10-25% in a bull market. Institutions commonly use it as a low-leverage stable-yield strategy.
5. Market Making. Both DEXs (Uniswap v3, Curve) and CEXs (Coinbase, Kraken) offer market-making opportunities. LP yield + liquidity incentives. Wintermute, GSR, Flow, B2C2, and Cumberland are the leading market makers.
6. Stat Arb · crypto statistical arbitrage. Long-short hedging across BTC / ETH / SOL / mid- and small-cap coins. Sharpe has broadly compressed to 1-2 post-2023.

Scale and players

⚠ Risk specific to crypto quant · Counterparty risk > market risk.

The 2022 collapses of FTX / Three Arrows / Alameda wiped out a swath of hedge funds and quant teams overnight. Your custody risk, exchange bankruptcy risk, and stablecoin depeg risk vastly exceed the market’s beta risk. The number one skill in crypto quant is fund routing + counterparty management, not alpha.

Hardware / Data / Talent — infra comparison across the eight schools

The table below is required reading before starting a quant business or applying for a quant job — the school you choose dictates the minimum resources you need to start. Mismatching school and resources is the deadliest mistake beginners make.

Practical takeaway · What should a two-person team with $100K do?

• Factor / Smart Beta (viable) — wrapped as mutual fund / SMA, 1-3% annualized excess return is achievable
• CTA trend (viable) — futures account + a database + basic rules is enough
• Crypto quant (viable but high-risk) — funding rate + simple arbitrage works, but custody risk is large
• Event-driven (viable) — earnings-drift strategies are low-cost
• HFT / hard StatArb / ML — don’t attempt at this scale, you’ll lose at the hardware / data starting line

How to choose — one chart to pick your school

1. I’m an individual investor with $10K-$1M. Options: factor ETFs (MTUM / VLUE / QUAL / USMV) + low-cost trend-following ETFs (KMLM / DBMF). If you write code, try simple momentum + mean-reversion combos, but don’t expect Sharpe > 1. Prioritize avoiding big mistakes.
2. I’m a small team with $1-10M. Options: CTA trend + event-driven + crypto quant. Avoid HFT and deep-learning ML. Spend 1-2 years building data and backtesting infrastructure before going live.
3. I want to work in quant / get a quant job. Foundational math (probability / statistics / linear algebra) + programming (C++ / Python) + deep mastery of one specialty (stochastic processes / econometrics / ML). Jane Street / Citadel / Two Sigma SWE/Quant comp starts at $300-500K, but it’s hyper-competitive.
4. I’m a large institution (bank / broker / insurer). Directly procure mature factor products (AQR / DFA) + a modest CTA allocation. Building an in-house quant team is uneconomic below $500M of capital — people + data costs exceed expected excess return.
5. I’m a founder with $10-50M seed. Pick one small but deep school (e.g. crypto funding rate, a niche category’s trend, the Quality tilt within factor), put up a 3-year live track record, then expand laterally. The biggest trap is wanting to do everything → doing nothing well.
6. I have $100M+ to allocate. Diversify across 3-4 schools and managers. CTA 20% + StatArb 30% + multi-strat platform 30% + factor 20% is a common configuration. The focus is “manager selection” rather than “strategy selection” — quant alpha comes mainly from people, not the school.

One-liner · Quant is a “resource-fit” game, not an “IQ game.”

Medallion looks like the smartest people winning, but underneath it’s the earliest (1988), the smallest (internal only), and the deepest (30 years of compounding) group locking in their seats. Replicating it today is near-impossible — but every scale has a suitable school. Picking the right play for your scale is ten times more important than chasing “Renaissance-level Sharpe.”

References — representative firms across the eight schools

High-frequency market making and execution (HFT)

• Citadel Securities — the largest US retail order-flow market maker. citadelsecurities.com
• Virtu Financial — publicly listed global electronic market maker. virtu.com
• Jump Trading — multi-asset HFT and crypto market making. jumptrading.com
• Jane Street — ETF authorized participant; options / Treasury market making. janestreet.com
• Hudson River Trading (HRT) — multi-strategy quant market making. hudsonrivertrading.com
• DRW — futures / derivatives market making. drw.com
• IMC — options market making and global electronic trading. imc.com
• Tower Research Capital — global electronic market making. tower-research.com

Mid-frequency StatArb / quantitative hedge

• Renaissance Technologies — Medallion fund, the quant myth. rentec.com
• D. E. Shaw — veteran quant hedge fund. deshaw.com
• Two Sigma — data-driven multi-strategy fund. twosigma.com
• Morgan Stanley — origin of PDT; veteran institutional quant research. morganstanley.com
• Citadel (Wellington / Tactical) — multi-strategy hedge fund flagship. citadel.com
• PDT Partners — statistical arbitrage boutique spun out of Morgan Stanley. pdtpartners.com
• WorldQuant — exemplar of the “alpha factory” model. worldquant.com

CTA / Trend following

• Man AHL — classic CTA / systematic trend. man.com/ahl
• Winton Group — systematic global macro. winton.com
• Aspect Capital — multi-horizon trend following. aspectcapital.com
• DUNN Capital — veteran CTA. dunncapital.com
• Graham Global — systematic macro and trend. grahamcapital.com
• Millburn — pioneer of systematic futures trading. millburn.com
• Transtrend — Dutch systematic CTA. transtrend.com
• Lynx Asset Management — Nordic systematic trend. lynxhedge.se

Risk Parity / Factor Investing

• AQR Capital Management — factor investing / academically-driven institutional asset management. aqr.com
• Bridgewater · Pure Alpha — All Weather / Pure Alpha flagships. bridgewater.com
• Dimensional Fund Advisors (DFA) — Fama-French factor incarnate. dimensional.com
• Research Affiliates — inventor of fundamental indexing. researchaffiliates.com
• BlackRock Systematic — systematic arm of the world’s largest asset manager. blackrock.com
• State Street Advanced Research — smart beta and factor products. ssga.com
• iShares · Vanguard Factor ETFs — factor ETF product suites. ishares.com

Machine learning / alternative data

• Voloridge Investment Management — flagship ML-first fund. voloridge.com
• Numerai — crowdsourced model hedge fund. numer.ai
• QuantConnect — quant research and backtesting platform. quantconnect.com

Multi-strategy platforms

• Millennium Management — pioneer of the multi-PM platform model. mlp.com
• Point72 — Steve Cohen’s multi-strategy platform. point72.com
• Balyasny Asset Management (BAM) — multi-strategy platform. bam.com
• Schonfeld Strategic Advisors — multi-strategy platform. schonfeld.com

Crypto market making

• Wintermute — leading crypto market maker. wintermute.com
• GSR — crypto market making and derivatives. gsr.io
• B2C2 — institutional crypto OTC. b2c2.com
• Cumberland (DRW) — crypto OTC and market making. cumberland.io