Unit 1 · The Quant Landscape — Session 01 of 05

What Quants Actually Do

"Quant" is not one job. It is at least five jobs, spread across half a dozen kinds of firm that make money in completely different ways: a market maker earns a sliver on a million trades, a trend follower makes a few big bets a year, a bank prices products for clients, an asset manager charges a small fee on a very large pile. Each business has its own P&L shape, its own way of dying, and its own idea of a good quant. The core move: follow the money — who pays the firm, for what, and what can kill it — and the job description follows.

DURATION 1.5 h + lab LEVEL Foundations TRACKS QR · QT · QD · IND HANDS-ON Two toy quant businesses, one market
📓 Companion lab Open in Colab Notebook on GitHub Source (.py)
On this page
  1. The four jobs in every quant business
  2. How an idea becomes a position
  3. The map: buy side, sell side & the firms between
  4. Follow the money: business models & fees
  5. Two businesses, one market
  6. Thin edges & lucky paths
  7. Choosing your game
  8. Try it
  9. Glossary · Check yourself · Go deeper

1.1.1 The four jobs in every quant business

Strip away the firm names and every quant business has to answer the same four questions, every day. The job titles are just who answers which one.

A quant uses mathematics, statistics and code to make trading and investment decisions, or builds the machinery that makes them. The forecaster, the person sizing the bet, the person whose code sends the orders and the person whose job is to say "no" are all quants. They sit around one thing: a book, the set of positions a firm holds, with its P&L (profit and loss) and its risk.

FOUR QUESTIONS, FOUR JOBS — and one book of positions they all serve QUANT RESEARCHER “What will happen?” signals · forecasts · models judged on: forecast quality and the P&L it leads to THE BOOK positions · P&L · risk owned by a portfolio manager QUANT TRADER “What do we do, how big?” sizing · execution · hedging judged on: P&L, Sharpe, drawdown and costs forecasts orders RISK & MODEL VALIDATION “What could kill us?” limits · stress tests · independent review of every model limits QUANT DEVELOPER “Can the machine do it — correctly, fast, every day?” data · backtester · trading platform · judged on uptime, latency, correctness An independent trader answers all four questions alone — the hardest seat there is.
fig 1.1.1 — the four questions every quant business answers, and the seat that answers each; the developer's platform sits under all of them

A quant researcher (QR) asks what will happen: they turn data and hypotheses into a signal — a number that forecasts something — and prove, as honestly as they can, that it works. A quant trader (QT) asks what do we do, and how big: sizing, execution, hedging, and living with the P&L. A quant developer (QD) asks can the machine do it: data pipelines, backtester, trading platform. Around them sit risk and model-validation quants, who ask what could kill us and can stop things, and the portfolio manager (PM), who owns the book, its capital and the final decision.

SeatA typical dayJudged onCourse track
Quant researchertesting hypotheses, writing memos, killing most ideasout-of-sample forecast quality; the P&L, Sharpe ratio and capacity of what shipsQR
Quant traderrunning live strategies, sizing, handling exceptionsP&L, Sharpe or information ratio, drawdown, execution costQT
Quant developerbuilding pipelines, the backtester, the order systemcorrectness, uptime, latencyQD
Risk / model-validation quantchallenging models, setting limits, stress testscatching problems before they cost money—
Portfolio managerdeciding what runs, how big, and what stopsnet P&L against risk and capital usedQR / QT
Independent systematic traderall of the above, alone, with their own moneyafter-cost, after-tax growth — and survivalIND
📌 Key idea — a role is the number you answer for

Titles vary wildly between firms; accountability does not. The researcher answers for whether the forecast was right, the trader for the P&L, the developer for whether the system did what it was told, and the risk quant for the loss nobody saw coming. When you read a job advert, ignore the title and ask which number you will be judged on. ("Quant" at a bank often means a pricing-model specialist; "quant trader" at an electronic firm, someone who runs automated strategies.)

1.1.2 How an idea becomes a position

Roles only make sense in motion. Follow one idea from a researcher's notebook to a live position and you meet every seat in the firm — and every place things go wrong.

ONE IDEA'S JOURNEY THROUGH THE FIRM — who holds it at each step 1 · RESEARCH quant researcher signal + honest backtest 2 · DECIDE PM / quant trader fits the book? how big? 3 · BUILD quant developer research code → production code 4 · SIGN OFF risk quant limits, stress tests, validation 5 · TRADE trader + algos orders, fills, real costs 6 · MONITOR everyone live vs expected, kill rule ready decay, surprises and bugs become the next questions Every arrow is a handoff, and every handoff is a place where research and reality drift apart.
fig 1.1.2 — one idea's path through the seats of a quant firm, with the monitoring loop that feeds the next idea

At a large firm each box is a different team; on your own you are every box, and the handoffs become checklists that are easy to skip. Session 1.3 — The Quant Research Lifecycle turns the top half of this picture into a research process with explicit gates; Unit 15 builds the bottom half as software.

HandoffWhat is handed overThe classic failure
Research → decidea backtest and a research memothe best of many accidental results, presented as a discovery
Decide → builda specification: universe, signal, sizing, limitsproduction code that quietly differs from the research code
Build → sign offa running system with its risk limits wired inlimits that live in a spreadsheet, not in the order path
Sign off → tradeapproval, capital and a kill switchtransaction costs larger than the backtest assumed
Trade → monitorlive P&L and fills, compared with expectationsnobody owns the decision to switch it off
💡 Looking ahead

Engineers tend to underrate step 3 and overrate step 1. A mediocre signal run by a correct, well-monitored system usually beats a brilliant one run by a leaky system, which is why the course spends as long on data (Unit 4), honest backtesting (Unit 6) and production systems (Unit 15) as on forecasting.

1.1.3 The map: buy side, sell side & the firms between

Every trade has two sides. The industry is organised around which side you are on, whose money you use, and who pays you for what.

THE MARKET ECOSYSTEM — whose money, whose orders, who gets paid END INVESTORS own the money pension funds endowments insurers sovereign funds individuals BUY SIDE invests the money asset managers index · factor · active hedge funds single-strategy · systematic multi-manager “pods” many PMs, one platform SELL SIDE sells services to it execution & broking gets orders done client market making flow desks quote clients derivatives & pricing structuring · XVA quants VENUES where trades meet exchanges other lit and dark venues clearing houses settle what trades capital returns orders fees orders PROP FIRMS & ELECTRONIC MARKET MAKERS own capital only · quote both sides on every venue paid by the spread and by exchange rebates quotes, all day orders money Every arrow that carries money is somebody's business model — §1.1.4 follows each one.
fig 1.1.3 — the ecosystem: end investors hand capital to the buy side, which sends orders through the sell side to venues, where prop firms and market makers quote all day with their own money

The buy side invests capital and buys trading services to do it: asset managers, hedge funds, and the pensions, endowments and insurers whose money most of them manage. The sell side — mostly banks and brokers — sells those services: execution, research, financing, derivatives, a price whenever a client asks. Orders flow from the buy side through the sell side to exchanges and other venues; money flows back as fees, commissions and spreads. Session 2.1 — Market Anatomy & the Life of an Order follows one order through that plumbing.

Some firms fit neither label. A proprietary trading firm trades only its own capital: it manages no client money (though it may have shareholders, even a listing), keeps every dollar it makes and eats every dollar it loses. Many of the best known are market makers, firms that continuously quote a price to buy and a price to sell on every venue that will have them. The table names a few well-known firms of each type, as of 2026, purely for orientation.

Firm typeWhose moneyWhat quants do thereExamples (as of 2026)
Proprietary trading firm / electronic market makerits ownshort-horizon signals, quoting, latency-critical systemsJane Street, Citadel Securities, Optiver, Hudson River Trading, IMC, Virtu
Systematic hedge fundclients' (and partners')large teams building many signals; portfolio constructionRenaissance Technologies, Two Sigma, D. E. Shaw, AQR, Man Group
Multi-manager ("pod") hedge fundclients'quants inside dozens of PM teams, plus central riskMillennium, Citadel, Balyasny, Point72, ExodusPoint
Sell-side bankthe bank's balance sheet, for clientspricing, structuring, XVA, flow trading, model validationthe large global banks' markets divisions
Asset managerclients'index and factor products, risk models, portfolio optimisationBlackRock, Vanguard, Dimensional
Crypto trading firmusually its ownmarket making and arbitrage across dozens of 24/7 venuescrypto-native market makers and the crypto desks of traditional firms

Engineers often overlook the sell side. Banks split trading into flow — standardised, high-volume products such as government bonds, currencies and plain swaps, priced fast and automatically for a small spread — and exotics, bespoke structured products sold at a higher margin, where the risk is a wrong pricing model. Bank quants build and hedge those models (Unit 10) and compute XVA, the valuation adjustments for counterparty credit, funding and capital.

🧭 Margin note — why prop trading left the banks (US, as of 2026)

The Volcker Rule — section 619 of the Dodd-Frank Act, with compliance required from 21 July 2015 and simplified by amendments in 2019–2020 — bars US banks from most short-term proprietary trading while exempting market making for clients and hedging, which is one reason prop trading now lives mainly in independent firms.

1.1.4 Follow the money: business models & fees

One question explains more about a quant job than any title: where does the firm's revenue come from, and what does it have to risk to earn it?

SEVEN WAYS TO GET PAID — the colour shows whose capital is at risk own capital clients' capital bank balance sheet, used for clients BUSINESS WHO PAYS IT WHAT CAN KILL IT ELECTRONIC MARKET MAKER own capital · quotes both sides impatient traders pay the spread; venues pay rebates for resting quotes informed traders (adverse selection), inventory, a technology failure PROP TRADING FIRM own capital · many strategies no clients: the P&L of its own bets is its only revenue every loss is its own — no client money to share it HEDGE FUND clients' capital · long and short investors: ≈1.3% of assets a year + ≈16% of profits (HFR averages) drawdowns make investors redeem; no performance fee below the old peak MULTI-MANAGER “POD” FUND clients' capital · dozens of PM teams investors: costs passed through + a performance fee each PM has a tight loss limit; breach it and the capital is pulled ASSET MANAGER clients' capital · mostly long-only investors: a small % of assets — 0.14% a year, average US index equity ETF tracking error; fee pressure from ever-cheaper rivals SELL-SIDE BANK balance sheet · client-driven clients: spreads on flow trades, fees on structured products wrong models, unhedged inventory, capital charges set by regulators CRYPTO TRADING FIRM own capital · 24/7 venues spreads, funding payments on perpetuals, gaps between venues a venue failing with the firm's money on it (FTX, November 2022)
fig 1.1.4 — seven business models: who pays each one, whose capital is at risk, and what can kill it (fee levels: HFR industry averages, 1Q 2026; ICI US averages, 2025)

Firms that keep their own P&L

A market maker is paid by people who want to trade now: it keeps the spread between its bid and offer, often collects a rebate for posting quotes (maker-taker fees), and risks trading with someone who knows more — adverse selection, formalised by Glosten & Milgrom (1985). Done well, its P&L is astonishingly smooth: Virtu's 2014 IPO filing reported one losing trading day in 1,238 (2009–2013). Menkveld (2013) found both halves of the business in one large high-frequency market maker: about €1.55 per trade earned on the spread, net of fees, and €0.68 lost on the positions left behind (figures from the 2011 working-paper version). The scale now rivals banks: Jane Street's 2024 net trading revenue, $20.5 billion, exceeded Citigroup's and Bank of America's trading revenue, according to documents seen by Bloomberg (April 2025).

Firms paid by investors

A hedge fund — a private fund for professional investors that can go long and short and use leverage — charges a management fee on assets and a performance fee on profits, the latter only above its high-water mark, the highest value the investor's stake has reached. The headline is "2 and 20"; HFR's industry averages for the first quarter of 2026 were 1.32% and 15.78%. Multi-manager funds ("pod shops") run dozens of PM teams, each with its own capital and tight loss limits, and most bill investors a pass-through of the fund's costs, staff pay included, plus a performance fee. An asset manager charges a small percentage of a very large pile — in 2025 the asset-weighted average was 0.14% a year for US index equity ETFs and 0.40% for US equity mutual funds (ICI) — so its quants build index, factor and risk products where precision and cost matter most.

In words: the investor gets the gross return minus the management fee, minus costs passed through, minus a share of whatever is left if it is positive. With gross return RgrossR_{\text{gross}}, management fee mm, pass-through costs cc and performance-fee rate pp, for one year that starts at the high-water mark:

Rnet=Rgross−m−c−pmax⁡(0,Rgross−m−c)R_{\text{net}} = R_{\text{gross}} - m - c - p\,\max\big(0,\; R_{\text{gross}} - m - c\big)
A +10% GROSS YEAR: WHAT THE INVESTOR KEEPS AND WHAT FEES AND COSTS TAKE investor keeps fees & costs 0.00% 2.50% 5.00% 7.50% 10.00% 12.50% US index ETF · investor keeps: 9.86% 9.86% US index ETF · fees & costs: 0.14% 0.14% US index ETF US mutual fund · investor keeps: 9.60% 9.60% US mutual fund · fees & costs: 0.40% 0.40% US mutual fund HF average · investor keeps: 7.31% 7.31% HF average · fees & costs: 2.69% 2.69% HF average HF 2 and 20 · investor keeps: 6.40% 6.40% HF 2 and 20 · fees & costs: 3.60% 3.60% HF 2 and 20 pod, illustrative · investor keeps: 3.20% 3.20% pod, illustrative · fees & costs: 6.80% 6.80% pod, illustrative
fig 1.1.5 — the same +10% gross year under five fee models: the business model, not the skill, decides how much reaches the investor
Fee modelTerms (per year)Investor, +10% gross yearInvestor, +4% gross year
US index equity ETF (average)0.14% of assets9.86%3.86%
US equity mutual fund (average)0.40% of assets9.60%3.60%
Hedge fund, industry average1.32% + 15.78% of profits7.31%2.26%
Hedge fund, "2 and 20"2% + 20% of profits6.40%1.60%
Multi-manager pass-through (illustrative)6% costs + 20% of profits3.20%−2.00%

The pass-through row is illustrative, but the pattern is documented: Bloomberg reported that at Balyasny a 15.2% gross return in 2023 became a 2.8% gain for investors, and a BNP Paribas estimate put investors' share of multi-strategy profits at 41 cents per dollar in 2023, down from 54 in 2021. Defenders reply that net returns stay competitive — a 2022 Barclays study of 300 funds found pass-through funds returned 11.8% after fees against 6.4% for peers (both estimates via eFinancialCareers) — so treat it as contested.

⚠️ Pitfall — "the fund made 10%, so its investors made 10%"

Only the net return reaches the investor, and the gap is set by the business model, not by skill. Ask of every performance figure: gross or net, net of which costs, and measured from where relative to the high-water mark. Ask the same of your own backtests: a strategy that only works before realistic costs does not work (Session 6.5).

Bank quants, finally, are judged on whether prices and hedges hold up, because clients, not market direction, pay the bank. Crypto trading firms live off spreads, funding payments on perpetual futures and gaps between venues, and carry a risk traditional firms hand to clearing houses: the venue itself can fail with your money on it — counterparty risk made real by FTX's collapse in November 2022 (Session 2.6).

1.1.5 Two businesses, one market

Business models are easy to describe and hard to feel, so the lab builds two tiny quant businesses and lets them trade the same market with the same money for ten years.

The lab simulates one liquid instrument for 10 years — annualised volatility 16%, no built-in risk premium — as a slow hidden drift that wanders up and down (the trends), plus noise, plus news jumps. Each business gets $1,000,000 and keeps trading that amount; profits are not reinvested.

  • The market maker quotes 2 basis points either side of the mid-price, earns a 0.2 bp rebate, and trades with about 400 clients a day, $10,000 each. 5% of them are informed — they know where the price is about to go — and on news days more arrive.
  • The trend follower is long when a fast moving average of the price is above a slow one and short otherwise, sizes to 15% annual volatility (at most 3× capital), and pays 5 bp per dollar traded.

The core of each toy is a few lines:

# market maker: every trade earns half the spread plus the rebate;
# an informed client takes back the move they knew was coming
edge = HALF_SPREAD + REBATE
informed = flow["u"] < informed_share
per_trade = np.where(informed, edge - flow["informed_move"], edge - flow["side"] * flow["hold_move"])

# trend follower: direction from two moving averages, size from volatility
direction = np.sign(log_price.ewm(span=FAST_SPAN).mean() - log_price.ewm(span=SLOW_SPAN).mean())
vol_est = market["ret"].ewm(span=VOL_SPAN).std() * np.sqrt(PERIODS_PER_YEAR)
weight = (direction * TARGET_VOL / vol_est).clip(-MAX_LEVERAGE, MAX_LEVERAGE).fillna(0.0)
held = weight.shift(1).fillna(0.0)                      # decided at yesterday's close
CUMULATIVE P&L AS A SHARE OF CAPITAL, BY YEAR ($1M EACH, SAME MARKET) market maker trend follower −50% 0% 50% 100% 150% 0 2 4 6 8 10 market maker market maker: 126% trend follower trend follower: 73% MM: 126% TF: 73%
fig 1.1.6 — cumulative P&L over ten simulated years: a near-straight ramp against a rollercoaster

Look at the shapes before the numbers. The market maker's curve is almost a straight line with small dents on loud news days ($126,154 a year on this path). The trend follower's lurches — long flat or falling stretches, then a few trends that make the decade ($72,894 a year) — while $1 simply held in the market became $1.02 (the toy market has no built-in risk premium). The daily P&L shows why:

MARKET MAKER: DAILY P&L (DAYS BELOW −$3K DRAWN AT THE LEFT EDGE) −$3.0k −$2.0k −$1.0k $0 $1.0k −$3.0k to −$2.9k −$2.9k to −$2.8k −$2.8k to −$2.7k −$2.4k to −$2.3k −$2.2k to −$2.1k −$2.1k to −$2.0k −$2.0k to −$1.9k −$1.9k to −$1.8k −$1.8k to −$1.7k −$1.7k to −$1.6k −$1.6k to −$1.5k −$1.5k to −$1.4k −$1.4k to −$1.3k −$1.3k to −$1.2k −$1.2k to −$1.1k −$1.1k to −$1.0k −$1.0k to −$900 −$900 to −$800 −$800 to −$700 −$700 to −$600 −$600 to −$500 −$500 to −$400 −$400 to −$300 −$300 to −$200 −$200 to −$100 −$100 to $0 $0 to $100 $100 to $200 $200 to $300 $300 to $400 $400 to $500 $500 to $600 $600 to $700 $700 to $800 $800 to $900 $900 to $1.0k $1.0k to $1.1k losing days ←
fig 1.1.7 — the market maker's daily P&L: a tall pile of small wins and a thin coral tail of losing days that stretches far to the left
TREND FOLLOWER: DAILY P&L (DAYS BEYOND ±$40K DRAWN AT THE EDGES) −$40.0k −$20.0k $0 $20.0k $40.0k −$40.0k to −$38.0k −$38.0k to −$36.0k −$36.0k to −$34.0k −$34.0k to −$32.0k −$32.0k to −$30.0k −$30.0k to −$28.0k −$28.0k to −$26.0k −$26.0k to −$24.0k −$24.0k to −$22.0k −$22.0k to −$20.0k −$20.0k to −$18.0k −$18.0k to −$16.0k −$16.0k to −$14.0k −$14.0k to −$12.0k −$12.0k to −$10.0k −$10.0k to −$8.0k −$8.0k to −$6.0k −$6.0k to −$4.0k −$4.0k to −$2.0k −$2.0k to $0 $0 to $2.0k $2.0k to $4.0k $4.0k to $6.0k $6.0k to $8.0k $8.0k to $10.0k $10.0k to $12.0k $12.0k to $14.0k $14.0k to $16.0k $16.0k to $18.0k $18.0k to $20.0k $20.0k to $22.0k $22.0k to $24.0k $24.0k to $26.0k $26.0k to $28.0k $28.0k to $30.0k $30.0k to $32.0k $32.0k to $34.0k $34.0k to $36.0k $36.0k to $38.0k $38.0k to $40.0k losing days ←
fig 1.1.8 — the trend follower's daily P&L: roughly half the days lose money, and both tails are wide

One path is one draw, so the table reports the median over 200 fresh ten-year worlds; the charts above show one of them.

Typical value over 200 worldsMarket makerTrend followerWhat it tells you
P&L per year on $1m$122,847$98,796same order of magnitude
Sharpe ratio (risk-free rate 0 in the toy)10.80.63wildly different reliability
Days that make money92%52%MM: most days; TF: a coin flip
Trades per year100,8052.9tiny bets vs a few large ones
Trades that make money65%31%hit rate alone says little…
Average win ÷ average loss1.35.1…without the payoff ratio
Skewness of daily P&L−7.90.01the MM's rare losing days are huge
Skewness of P&L per trade−0.32.7the TF's few winners are huge
Worst day, in days of average profit26119how long one bad day takes to recoup
Maximum drawdown (share of capital)−1.3%−39%the pain investors sit through
Longest time below a previous peak30 trading days762 trading daysweeks vs years

The market maker is a classic negative-skew business. The trend follower is its mirror image: it loses most trades but wins several times as much when it wins — on this path its three best trades earned 60% of its gross gains, the option-like profile Fung & Hsieh (2001) found in real trend-following funds. Their daily P&L is typically uncorrelated (median correlation 0.00), which is the logic behind multi-strategy firms.

📌 Key idea — the business model sets the return shape

A market maker sells immediacy and is paid a little on almost every trade; a trend follower bets that moves continue and is paid a lot on a few. The first wins most days and occasionally loses badly; the second loses most trades and occasionally wins big. Know the shape of your P&L before you judge a single number in it — a bad month means something different in each business, and so do the risk limits.

1.1.6 Thin edges & lucky paths

A typical Sharpe ratio of 11 looks like free money, and a trend follower that typically spends 762 trading days under water looks broken. Neither impression survives a closer look — and the reasons are what quants in each business spend their days on.

The market maker's edge is thinner than it looks

In words: every everyday trade earns the half-spread hh plus the rebate bb; a fraction qq of clients are informed and each takes back the move they knew was coming, |Δ|\lvert\Delta\rvert on average; and news days cost a roughly fixed ℓ\ell per everyday trade. So the expected P&L per trade, and the break-even informed share, are

𝔼[P&L per trade]=(h+b)−q𝔼|Δ|−ℓ,q⋆=h+b−ℓ𝔼|Δ|\E[\text{P\&L per trade}] = (h + b) - q\,\E\lvert\Delta\rvert - \ell, \qquad q^{\star} = \frac{h + b - \ell}{\E\lvert\Delta\rvert}

Here h+bh + b is 2.2 bp and 𝔼|Δ|\E\lvert\Delta\rvert is 7.98 bp. Without news (ℓ=0\ell = 0) the business would break even at 27.6% informed clients; with the measured ℓ\ell of 0.55 bp the formula predicts 20.7% and the simulation finds 20.7%. The lab keeps the client flow fixed and turns up only the informed share:

MARKET MAKER'S ANNUAL P&L AS THE SHARE OF INFORMED CLIENTS RISES −$100k $0 $100k $200k 0%: $166k $166k 0% 5%: $126k $126k 5% 10%: $86k $86k 10% 15%: $46k $46k 15% 20%: $6k $6k 20% 25%: −$35k −$35k 25%
fig 1.1.9 — the market maker's annual P&L as the share of informed clients rises: same flow, same spread, and the edge is gone before a quarter of clients know more than you

At 5% informed the toy earns $126,154 a year; at 15% it earns $45,984; at 25% it loses $34,790. That is adverse selection: the spread is the fee for trading with you, and it has to pay for the clients who know more than you do. It is what market-making quants obsess over — widening quotes before news, cancelling stale quotes fast (why latency matters), managing inventory risk, and prizing flow they believe is uninformed. Session 14.2 — Microstructure Models gives the models; Session 14.5 — Market Making builds a real quoting strategy.

⚠️ Pitfall — "a Sharpe ratio of 11? Market making must be the best business"

The toy leaves out what makes the edge expensive — the technology race, fees, colocation, staff, and competitors who narrow the spread until the marginal firm barely breaks even — and hides the capacity limit: profit comes from client flow, not capital, so twice the money earns the same dollars. Menkveld's market maker had an annualised Sharpe ratio of about 9 in the 2011 working-paper version; the published version stresses that the figure depends heavily on how much capital you assume it ties up. A high Sharpe says the edge is reliable, not that it is available to you.

The trend follower's edge takes decades to prove

Across the 200 worlds, the market maker's Sharpe ratio lands between 8.8 and 13.0 in 80% of them and is never negative. The trend follower — whose edge is real by construction — lands anywhere between 0.04 and 1.27 and loses money over a whole decade in 7.5% of worlds. The paths in the charts are ordinary draws, at percentiles 33 (market maker) and 34 (trend follower).

This changes how quants are judged. A market-making team can test a change to its quoting in days, because 100,805 trades a year make even a tiny difference visible. A trend-following team cannot: with a few trades a year its P&L is mostly luck over any horizon a manager has patience for, so it is judged on process and on evidence pooled across many markets and decades — which is where the real case for trend following comes from (Session 8.1 — Time-Series Momentum & Trend Following). How many years a Sharpe ratio needs before you can believe it is Session 5.2's question; Session 1.2 — Where Returns Come From names each toy's source of edge.

1.1.7 Choosing your game

You know the seats and the businesses. Now match them to what you are good at, what you enjoy, and what you can realistically get into.

WHICH GAME ARE YOU TRAINING FOR? — a first sort, not a verdict YOU, AFTER THIS COURSE WHOSE CAPITAL? a firm's, or yours? my own INDEPENDENT (IND) slow strategies, small size; you are all four seats · Unit 16 a firm's WHAT DO YOU WANT TO OWN? the part you would do for free QUANT RESEARCHER you want to own the forecast Units 5–8, 12–13 QUANT TRADER you want to own the P&L Units 9 and 14 QUANT DEVELOPER you want to own the machine Units 4, 6 and 15 RISK / VALIDATION you want to own the veto Units 9–11 Careers cross these lines, and every seat needs the others' language — so the course teaches all four.
fig 1.1.10 — a first sort of the four tracks: whose capital you want to use, then which number you want to answer for

Match the game to your strengths. Love fast, correct systems? Electronic market making and quant development. Love finding structure in noisy data and proving it honestly? Research at a systematic fund. Like deciding under pressure? Trading. Instinct to break other people's models? Risk and validation. And if you want to trade your own money, pick games where being small is not a handicap — slower strategies with modest turnover — because §1.1.6 showed why you cannot win the market maker's game from a laptop: speed and flow decide your informed share.

Be clear-eyed about the funnel, too: the famous seats are few — one veteran buy-side quant, Giuseppe Paleologo, informally estimated in 2024 that there are about 700 new buy-side quant jobs a year worldwide. Session 16.1 — Career Paths & the Quant Job Market covers the job market properly, Session 16.4 — Running Your Own Capital covers trading your own book, and Session 1.4 — The Engineer's Edge, Blind Spots & Honest Odds gives the evidence on who actually makes money.

⚠️ Pitfall — "I'll aim for whichever seat pays most"

The highest-paid seats are paid out of P&L, and P&L-linked seats are the least forgiving: a pod PM who breaches a loss limit can lose their capital in weeks. Pick the seat whose daily loop you would enjoy even in a bad year — the forecast, the P&L, the machine or the veto — because the job is mostly that loop, repeated for years.

1.1.8 Try it

Open the companion lab and run two small businesses. Predict before you run; that habit is the whole course in miniature.

🔧 Try it (15 min, in the lab)

1. A competitor halves the spread. Before running anything, use the formula in §1.1.6 to predict the break-even informed share with HALF_SPREAD at 1 bp instead of 2, ignoring news (ℓ=0\ell = 0). Then set HALF_SPREAD = 1.0e-4, rerun, and compare with breakeven_no_news. How far does the annual P&L fall?
2. Set it back, then double CAPITAL. Which business earns twice as many dollars, and which earns the same dollars on twice the money? Then double TARGET_VOL instead: what happens to the trend follower's Sharpe ratio, and to its maximum drawdown?
3. Success: your step-1 prediction is within a percentage point, and you can say in one sentence each why the market maker's profit does not scale with capital, and why the trend follower's Sharpe ratio barely changes with leverage while its drawdown doubles (the 3× cap binds in calm spells; see Session 2.3).

1.1.9 Glossary · Check yourself · Go deeper

Quant researcher — finds, tests and documents predictive signals; judged on out-of-sample forecast quality and the P&L it leads to.
Quant trader — turns forecasts into positions (sizing, execution, hedging); judged on P&L and risk-adjusted return.
Quant developer — builds the data, research and trading infrastructure; judged on correctness, uptime and speed.
Portfolio manager — the person accountable for a book's P&L, who decides what to hold and how much risk to take.
Buy side — firms that invest capital and buy trading services to do it: asset managers, hedge funds, pensions, insurers.
Sell side — firms that sell trading services to the buy side — execution, market making, research, financing, derivatives; mostly banks and brokers.
Proprietary trading firm — a firm that trades only its own capital — it manages no client money; it keeps all its P&L and bears all its losses.
Market maker — a firm that continuously quotes both a bid and an ask, earning the spread and bearing inventory and adverse-selection risk.
Hedge fund — a private fund for professional investors that can go long and short and use leverage; paid a management fee and a performance fee.
Asset manager — a firm that invests pooled or segregated client money, mostly long-only, for a fee that is a percentage of assets.
Management fee — an annual fee charged as a percentage of the assets managed, paid whatever the performance.
Performance fee — a share of the profits paid to the manager, usually only on gains above the high-water mark.
High-water mark — the highest value an investor's stake has reached; performance fees are paid only on gains above it.

Check yourself

1. A friend's new job is "quant at a bank, building pricing models for the exotic-options desk." Buy side or sell side — and what will they be judged on?
Sell side. Clients pay the bank for the products, so your friend is judged on whether prices and hedges hold up and survive independent validation — not on predicting where the market goes.
2. In the lab's world, a rival starts capturing the market maker's uninformed clients, so the informed share of what is left rises from 5% to 15%. What happens to annual P&L, and why?
It falls from about $126,154 to about $45,984, and break-even is near 21%. Each informed trade costs the move the client knew about, about 8.0 bp against 2.2 bp earned — so who you trade with matters more than how often. That is adverse selection.
3. A hedge fund makes +10% gross. What does its investor get under "2 and 20", and under 6% pass-through costs plus a 20% performance fee? And in a +4% year?
+10%: 6.40% and 3.20%. +4%: 1.60% and −2.00%. Pass-through costs are charged whether or not the fund makes money, so in a lean year the fund can be up while its clients are down.
4. A trend follower has spent two and a half years below its previous peak, and the boss wants to shut it down. What would you check before agreeing?
Whether it is inside what the strategy's own evidence calls normal: in the lab the median longest spell under water is 762 trading days (about three years), and a genuinely profitable trend follower still loses money over a whole decade in 7.5% of worlds. Judge the process: is it trading as specified, and are costs as modelled? A market maker that long under water would be broken — its typical worst spell is 30 trading days.
5. A trend follower wins only 31% of its trades. Is it a bad strategy?
Not on that evidence. Its average win is typically 5.1 times its average loss. A hit rate means nothing without the payoff ratio — the market maker wins 65% of its trades with the opposite shape.
🔧 Homework (~1 h before Session 1.2)

1. Pick three firm types from the table in §1.1.3. For each, write three lines: who pays it, what could kill it, and which seat you would want there. Keep the note for Session 16.1.
2. In the lab, add a third business: an index investor that holds the market every day and pays the 0.14% ETF fee, and run business_stats on it. On this path $1 held all along became $1.02; write one sentence on why this business has no edge in the toy market, then read Session 1.2 — Where Returns Come From with that sentence in mind.
3. Read chapters 1–3 of Harris (2003), below — a clear, practitioner-oriented map of who trades and why.

📚 Go deeper

  1. Harris, L. (2003). Trading and Exchanges: Market Microstructure for Practitioners, ch. 1–3 (who trades, and the trading industry). Oxford University Press. ISBN 978-0-19-514470-3
  2. Narang, R. K. (2013). Inside the Black Box: A Simple Guide to Quantitative and High-Frequency Trading, 2nd ed. Wiley. ISBN 978-1-118-36241-9
  3. Derman, E. (2004). My Life as a Quant: Reflections on Physics and Finance. Wiley. ISBN 978-0-471-39420-4
  4. Glosten, L. R. & Milgrom, P. R. (1985). "Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders." Journal of Financial Economics 14(1), 71–100. doi:10.1016/0304-405X(85)90044-3
  5. Menkveld, A. J. (2013). "High Frequency Trading and the New Market Makers." Journal of Financial Markets 16(4), 712–740. doi:10.1016/j.finmar.2013.06.006 — the per-trade figures and Sharpe ratio quoted here are from the 2011 working-paper version, SSRN 1722924
  6. Fung, W. & Hsieh, D. A. (2001). "The Risk in Hedge Fund Strategies: Theory and Evidence from Trend Followers." Review of Financial Studies 14(2), 313–341. doi:10.1093/rfs/14.2.313

Sources for the numbers on this page

  1. Virtu Financial, Inc. (2014). Form S-1 Registration Statement, filed 10 March 2014 (one losing day in 1,238, 2009–2013). https://www.sec.gov/Archives/edgar/data/0001592386/000104746914002070/a2218589zs-1.htm
  2. Bloomberg News (2025, 23 April). "Jane Street's $20.5 Billion Trading Haul Tops Citigroup, BofA." Bloomberg. https://www.bloomberg.com/news/articles/2025-04-23/jane-street-s-20-5-billion-trading-haul-tops-citigroup-bofa
  3. HFR (2026, 14 July). "Hedge Fund Launches, Liquidations Rise to Begin Volatile 2026." Press release (industry-average fees, 1Q 2026). https://www.hfr.com/media/market-commentary/hedge-fund-launches-liquidations-rise-to-begin-volatile-2026/
  4. Investment Company Institute (2026, 25 March). "Mutual Fund and ETF Fees Remained Near Historic Lows in 2025." News release (US, asset-weighted). https://www.ici.org/news-release/mutual-fund-and-etf-fees-remained-near-historic-lows-in-2025
  5. Parmar, H., Li, W., Kim, K. & Emamdjomeh, A. (2025, 11 February). "'No Limit' Hedge Fund Fees Are Eating Up Their Clients' Gains." Bloomberg. https://news.bloomberglaw.com/insurance/no-limit-hedge-fund-fees-are-eating-up-their-clients-gains
  6. Butcher, S. (2024, 4 November). "How hedge funds pay so highly: pass through fees, and their nuances." eFinancialCareers (BNP Paribas and Barclays figures). https://www.efinancialcareers.co.uk/news/hedge-fund-pass-throughs
  7. Butcher, S. (2024, 26 February). "World's wittiest quant advises how to get a hedge fund job." eFinancialCareers (Paleologo's estimate). https://www.efinancialcareers.be/en/news/hedge-fund-quant-job
  8. Congressional Research Service (2014, 27 March). "The Volcker Rule: A Legal Analysis." Report R43440 (the ban and its market-making and hedging exemptions). https://www.everycrsreport.com/reports/R43440.html
  9. Board of Governors of the Federal Reserve System. "Volcker Rule" (full conformance required by 21 July 2015). https://www.federalreserve.gov/supervisionreg/volcker-rule.htm
  10. US Securities and Exchange Commission (2019, 8 October). "Agencies Finalize Changes to Simplify Volcker Rule." Press release 2019-207. https://www.sec.gov/newsroom/press-releases/2019-207
← Course home Next: 1.2 · Sources of Return →

Educational material only — not investment, legal or tax advice. Backtests and simulations are hypothetical. Disclaimer · Text CC BY 4.0 · Code MIT.