Research Quantitative Finance

You'll learn

• Q-measure vs P-measure — risk-neutral pricing probabilities versus physical real-world forecasting probabilities
• Sell-side vs Buy-side quant — valuation and hedging of contracts versus capturing statistical risk premia and alpha
• Primary literature venues — Mathematical Finance, Journal of Financial Economics, Quantitative Finance, and arXiv q-fin

How to work through it

1. 1Outline the core differences between sell-side derivatives pricing and buy-side alpha generation.
2. 2Identify the defining journals and repositories for each branch (e.g., Mathematical Finance, Journal of Finance, SSRN, arXiv q-fin).
3. 3Draft a comparative reference table covering purpose, assumptions, and validation metrics for each branch.

You'll learn

• Paper extraction rubric — structured methodology for auditing financial claims, data selection, and proofs
• Implicit friction assumptions — zero transaction costs, infinite liquidity, continuous rebalancing, and unconstrained short-selling
• Identification strategy — how empirical researchers isolate causal effects from market noise

How to work through it

1. 1Create a standardized template covering: Thesis, Core Assumptions, Data Sources, Mathematical Framework, Econometric Identification, and Fragility Vectors.
2. 2Select a benchmark paper (e.g., Markowitz 1952 portfolio selection) to run through the template.
3. 3Identify every explicit and implicit friction assumption made in the paper.

You'll learn

• Sigma-algebra — mathematical collection of measurable events representing possible market states
• Filtration — increasing family of sigma-algebras representing information available up to time t
• Martingale property — conditional expectation of the next value given present information equals the present value

How to work through it

1. 1Define probability space triplet (Omega, F, P) and sigma-algebra properties.
2. 2Construct a filtration index representing discrete time steps.
3. 3Work through conditional expectation properties given a sub-sigma-algebra.
4. 4Prove the martingale, sub-martingale, and super-martingale conditions for discrete random walks.

You'll learn

• Wiener process — continuous-time stochastic process with independent, stationary Gaussian increments
• Quadratic variation — sum of squared differences over a partition, scaling linearly with time dt
• Itô's Lemma — the fundamental chain rule of stochastic calculus incorporating the second-order volatility correction

How to work through it

1. 1Define standard Brownian motion properties: stationary, independent Gaussian increments and continuous paths.
2. 2Compute the quadratic variation of Brownian motion and show why (dW)^2 = dt in the limit.
3. 3Taylor expand a function f(t, W_t) to second order and derive Itô's Lemma.
4. 4Apply Itô's Lemma to solve for Geometric Brownian Motion S_t = S_0 * exp((mu - 0.5 * sigma^2)*t + sigma * W_t).

You'll learn

• Ornstein-Uhlenbeck process — mean-reverting stochastic process widely used for interest rates and statistical arbitrage spreads
• Integrating factor method for SDEs — technique for solving linear stochastic differential equations
• Stationary distribution — long-term probability distribution of an ergodic stochastic process

How to work through it

1. 1Write the Ornstein-Uhlenbeck SDE: dX_t = theta * (mu - X_t) * dt + sigma * dW_t.
2. 2Use an integrating factor and Itô's Lemma to solve the OU SDE explicitly.
3. 3Compute the expected value, covariance function, and long-term asymptotic distribution of X_t.
4. 4Format the derivations into a concise, LaTeX-formatted or handwritten reference memo.

You'll learn

• Limit Order Book (LOB) — the real-time collection of unexecuted limit orders organized by price and arrival time
• Continuous Double Auction — market mechanism where buyers and sellers continuously submit bids and offers
• Adverse selection in trading — risk that resting limit orders are executed against better-informed market participants

How to work through it

1. 1Diagram the bid and ask sides of a limit order book across tick levels.
2. 2Contrast Price-Time priority (FIFO) with Size-Time (Pro-Rata) matching algorithms.
3. 3Analyze order-flow toxicity, spread crossing costs, and adverse selection incurred by resting passive orders.
4. 4Document how tick size constraints create queue congestion and affect volatility.

You'll learn

• Kyle's Lambda — measure of permanent price impact per unit of volume traded
• Glosten-Milgrom model — sequential trade model demonstrating that asymmetric information alone induces a positive bid-ask spread
• Price impact decomposition — separation of order flow effects into temporary (liquidity friction) and permanent (information) components

How to work through it

1. 1Read Kyle (1985) focusing on the three-player game: informed trader, noise traders, and competitive market maker.
2. 2Examine the derivation of Kyle's lambda (illiquidity parameter) and price impact.
3. 3Read Glosten-Milgrom (1985) and analyze the Bayesian updating rule used by the market maker upon observing buy vs sell order flow.
4. 4Write a side-by-side comparison of discrete (Glosten-Milgrom) versus continuous/batch (Kyle) approaches.

You'll learn

• Order Book Imbalance (OBI) — ratio of bid to ask volume at top levels predicting near-term price changes
• VPIN — Volume-Synchronized Probability of Toxicity estimating adverse selection risk in continuous time buckets
• Microstructure noise — high-frequency pricing artifacts caused by bid-ask bounce and discrete tick sizes

How to work through it

1. 1Study the construction of order book imbalance (OBI) across top-of-book levels.
2. 2Review the definition and calculation steps for VPIN (Easley, López de Prado, O'Hara, 2012).
3. 3Examine critiques and counter-arguments regarding parameter sensitivity and false alarm rates during regime shifts.
4. 4Synthesize findings into an evaluation memo outlining when order flow indicators provide true signal versus noise.

You'll learn

• Equivalent Martingale Measure (EMM / Q-measure) — probability measure under which discounted asset prices are martingales
• First Fundamental Theorem of Asset Pricing — equivalence between no-arbitrage and existence of a risk-neutral measure
• Girsanov's Theorem — mathematical theorem providing the change of measure formula eliminating drift in continuous processes

How to work through it

1. 1Study the 1st FTAP: No arbitrage is equivalent to the existence of at least one Equivalent Martingale Measure (EMM).
2. 2Study the 2nd FTAP: In an arbitrage-free market, completeness is equivalent to the uniqueness of the EMM.
3. 3Understand Girsanov's Theorem and how drift is adjusted from physical measure P to risk-neutral measure Q.
4. 4Define the Radon-Nikodym derivative dQ/dP and the market price of risk.

You'll learn

• Delta hedging — dynamically offsetting option exposure with the underlying asset to create a riskless portfolio
• Black-Scholes PDE — parabolic partial differential equation governing the no-arbitrage price of European contingent claims
• Put-Call Parity — static no-arbitrage relationship connecting European call and put prices, forward price, and strike

How to work through it

1. 1Construct a self-financing delta-hedged portfolio consisting of long underlying and short option.
2. 2Apply Itô's Lemma to the portfolio and show how the dW_t stochastic term is eliminated by choosing delta = df/dS.
3. 3Equate portfolio return to the risk-free rate r*dt to arrive at the Black-Scholes PDE.
4. 4Solve the PDE boundary value problem or integrate the risk-neutral discounted terminal payoff E_Q[exp(-rT)*max(S_T - K, 0)].
5. 5Verify Put-Call parity algebraically using the resulting formulas.

You'll learn

• Option Greeks — partial derivatives of option price with respect to underlying price, volatility, time, and interest rate
• Volatility Skew / Smile — empirical phenomenon where out-of-the-money puts trade at higher implied volatilities than ATM calls
• Local Volatility (Dupire) vs Stochastic Volatility (Heston) — deterministic state-dependent diffusion versus continuous two-factor stochastic variance

How to work through it

1. 1Calculate analytical formulas for Delta, Gamma, Vega, and Theta.
2. 2Graphically interpret Gamma-Theta trade-off in option holding.
3. 3Analyze market data anomalies: post-1987 stock crash volatility skew, fat-tailed returns, and jump risk.
4. 4Contrast Dupire's local volatility model with Heston's stochastic volatility model.
5. 5Synthesize your notes into an analytical critique of the volatility surface.

You'll learn

• CAPM — Capital Asset Pricing Model linking expected return strictly to systematic market beta
• Fama-French Multi-Factor Models — empirical asset pricing models adding Size, Value, Profitability, and Investment factors
• Fama-MacBeth two-pass regression — econometric procedure estimating factor risk premia across an asset universe

How to work through it

1. 1Study the CAPM single-factor formulation and its empirical shortcomings.
2. 2Read Fama & French (1992, 1993) on size (SMB) and value (HML) factors.
3. 3Analyze the Carhart (1997) momentum extension (UMD) and Fama-French 5-factor (RMW, CMA) additions.
4. 4Break down Fama-MacBeth methodology: time-series beta estimation followed by cross-sectional risk price estimation.

You'll learn

• Cross-sectional momentum — persistent outperformance of past winning assets over past losers over 3-12 month horizons
• Short-term reversal — microstructure- and liquidity-driven bounce back over 1-day to 1-month horizons
• Momentum crashes — violent drawdowns occurring during sudden market rebounds when short legs rally violently

How to work through it

1. 1Read Jegadeesh & Titman (1993) on cross-sectional 3-12 month momentum.
2. 2Examine short-term reversal (1-week to 1-month) and long-term mean reversion (3-5 years).
3. 3Compare behavioral explanations (investor underreaction, disposition effect) with risk-based models (time-varying risk premia).
4. 4Analyze momentum crash dynamics during sudden market regime turnarounds.

You'll learn

• Short rebate and borrow fees — financing cost paid to prime brokers to borrow shares for short sales
• Capacity constraint — maximum asset size a strategy can manage before market impact eliminates expected returns
• Factor crowding — concentration of capital in identical factors leading to liquidity crises during unwinds

How to work through it

1. 1Analyze transaction cost models: bid-ask spread, commissions, exchange fees, and market impact functions.
2. 2Investigate stock loan mechanics: hard-to-borrow fees, borrow recall risk, and short sale restrictions.
3. 3Evaluate factor crowding metrics: pairwise correlation spikes, institutional co-ownership, and fire-sale risk.
4. 4Draft a research synthesis memo detailing the gap between gross academic paper performance and net investable returns.

You'll learn

• Cointegration — linear combination of non-stationary time series that forms a stationary series
• Spurious regression — statistically invalid high R-squared relationships resulting from non-stationary trending variables
• Half-life of mean reversion — time required for an anomalous spread deviation to decay to half its magnitude

How to work through it

1. 1Define stationary (I(0)) versus integrated (I(1)) time series and unit root testing (Augmented Dickey-Fuller test).
2. 2Explain why regression on non-stationary series produces spurious correlation.
3. 3Derive the Engle-Granger two-step cointegration test.
4. 4Formulate the spread dynamics as an Ornstein-Uhlenbeck process and calculate half-life of mean reversion.

You'll learn

• Almgren-Chriss framework — standard optimal execution model balancing market impact against timing risk
• Temporary vs Permanent market impact — transient liquidity concession versus permanent revision of the equilibrium price
• TWAP & VWAP — Time-Weighted and Volume-Weighted Average Price algorithmic execution benchmarks

How to work through it

1. 1Read Almgren-Chriss (2000) focusing on the trade-off between price impact and timing risk.
2. 2Distinguish temporary price impact (instantaneous liquidity depletion) from permanent price impact (information update).
3. 3Formulate the objective function: minimize expected transaction cost plus risk penalty lambda * variance.
4. 4Solve the discrete Euler-Lagrange equations to find the optimal exponential/linear liquidation trajectory.

You'll learn

• Authorized Participant (AP) arbitrage — institutional mechanism keeping ETF market prices aligned with Net Asset Value (NAV)
• Cross-asset lead-lag — delayed information transmission from highly liquid macro instruments (futures/ETFs) to single stocks
• Price discovery transmission — sequence through which order flow in one market alters quotes in related markets

How to work through it

1. 1Examine ETF creation/redemption mechanics and Authorized Participant (AP) arbitrage.
2. 2Investigate empirical evidence on index futures leading cash equities during high-volatility events.
3. 3Analyze latency arbitrage and structural lead-lag caused by differing tick size constraints and participant mixes.
4. 4Synthesize your notes into an architectural diagram and memo on cross-asset price discovery.

You'll learn

• Factor Zoo — the proliferation of hundreds of empirical factors claiming to explain equity returns
• Multiple testing problem — mathematical surge in false discovery probability when running repeated statistical tests
• Benjamini-Hochberg procedure — statistical method controlling the false discovery rate across multiple simultaneous hypotheses

How to work through it

1. 1Read Harvey, Liu & Zhu (2016) documenting over 300 published 'factors' ('the factor zoo').
2. 2Study Family-Wise Error Rate (FWER) and Bonferroni correction.
3. 3Study False Discovery Rate (FDR) and Benjamini-Hochberg-Yekutieli adjustments.
4. 4Evaluate how p-hacking and publication bias skew academic and commercial finance literature.

You'll learn

• Backtest Overfitting — tuning strategy parameters until historical performance is maximized at the expense of out-of-sample validity
• Deflated Sharpe Ratio (DSR) — metric adjusting the observed Sharpe ratio for non-normality, sample length, and number of trials
• Selection Bias in Backtesting — reporting only the best-performing strategy iteration while hiding failed variants

How to work through it

1. 1Study how the maximum Sharpe ratio among N independent false strategies grows with sqrt(2 * ln(N)).
2. 2Derive the Probabilistic Sharpe Ratio (PSR) correcting for non-normal returns (skewness and kurtosis).
3. 3Derive the Deflated Sharpe Ratio (DSR) discounting for the number of backtest trials and selection bias.
4. 4Practice calculating DSR given trial history metadata.

You'll learn

• Point-in-time data — historical database capturing exactly what data was known at each historical timestamp, without post-event revisions
• Survivorship bias — systematic distortion caused by excluding companies that went bankrupt, merged, or delisted from historical backtests
• Lookahead bias — inadvertent inclusion of future information into trading signal computation

How to work through it

1. 1Document lookahead biases: using split/dividend adjusted data prematurely, end-of-day prices for intraday signals, and future aggregate data.
2. 2Document data biases: survivorship bias in stock universes, unadjusted index reconstitutions, and historical corporate fundamental restatements.
3. 3Document market friction biases: assuming fills at mid-price, ignoring minimum tick sizes, assuming infinite borrow at zero cost.
4. 4Assemble the items into an actionable checklist accompanied by forensic questions to ask the strategy designer.

You'll learn

• Coherent risk measure — risk metric satisfying monotonicity, subadditivity, positive homogeneity, and translation invariance
• Subadditivity — principle that the risk of a combined portfolio cannot exceed the sum of the risks of its constituent parts
• Expected Shortfall (CVaR) — average loss incurred given that a loss exceeds the Value at Risk threshold

How to work through it

1. 1Define the four axioms of coherent risk measures: Monotonicity, Subadditivity, Positive Homogeneity, and Translation Invariance.
2. 2Compute parametric VaR, Historical Simulation VaR, and Cornish-Fisher VaR.
3. 3Construct a counterexample where VaR(A + B) > VaR(A) + VaR(B).
4. 4Define Expected Shortfall (CVaR) as the conditional expectation of loss exceeding VaR and prove its subadditivity.

You'll learn

• Generalized Pareto Distribution (GPD) — continuous distribution modeling the excess loss over a high threshold in Extreme Value Theory
• Sklar's Theorem — theorem establishing that any multivariate joint distribution can be written in terms of univariate marginals and a copula
• Tail dependence coefficient — probability that one variable exceeds an extreme quantile given that another variable has exceeded it

How to work through it

1. 1Study the Fisher-Tippett-Gnedenko theorem and the Generalized Extreme Value (GEV) distribution.
2. 2Study the Pickands-Balkema-de Haan theorem and Peaks Over Threshold (POT) approach using Generalized Pareto Distribution (GPD).
3. 3Understand Sklar's Theorem for decomposing joint distributions into marginals and copula dependency structures.
4. 4Compare upper and lower tail dependence coefficients between Gaussian copula (zero tail dependence) and Student-t copula.
5. 5Document how the assumption of zero tail dependence contributed to the mispricing of structured credit tranche risk.

You'll learn

• Funding liquidity — availability of credit, margin, and repo financing to levered financial intermediaries
• Margin spiral — vicious cycle where increasing volatility causes lenders to raise margin requirements, forcing further asset sales
• Endogenous risk — financial risk generated internally by the collective actions and common risk management constraints of market participants

How to work through it

1. 1Distinguish between market liquidity (ease of trading an asset) and funding liquidity (ease of securing leverage/financing).
2. 2Map the two-way feedback: asset price drop -> margin calls -> forced liquidations -> further price drops.
3. 3Examine the loss spiral (capital depletion) and margin spiral (haircut escalation during volatility).
4. 4Synthesize findings into an institutional risk briefing document.

You'll learn

• Triple-Barrier Method — labelling financial time series based on whether price hits a stop-loss, profit-target, or time-expiration first
• Purged K-Fold Cross-Validation — cross-validation algorithm designed to eliminate label leakage caused by overlapping forward returns
• Fractional Differentiation — mathematical technique to achieve time series stationarity while preserving long-term memory

How to work through it

1. 1Examine the Triple-Barrier labelling method (upper profit-take, lower stop-loss, vertical time-barrier).
2. 2Understand why serial correlation and overlapping labels cause information leakage in standard K-Fold CV.
3. 3Mathematically define Purging (removing training labels that overlap with test labels) and Embargoing (removing training data immediately following test sets).
4. 4Study fractional differentiation: achieving stationarity without completely erasing multi-period memory.

You'll learn

• Signal-to-Noise Ratio (SNR) in finance — extreme dominance of stochastic noise over predictable deterministic drift in market returns
• Non-stationarity — constant statistical evolution of financial data generating mechanisms over time
• Inductive bias — assumptions built into an algorithm determining how well it generalizes to unseen market regimes

How to work through it

1. 1Examine the Signal-to-Noise Ratio (SNR) of financial returns compared to computer vision and natural language domains.
2. 2Review empirical comparisons between LightGBM/XGBoost and deep neural networks on tabular factor data.
3. 3Analyze the failure modes of Reinforcement Learning in non-stationary environments where transition probabilities shift abruptly.
4. 4Draft an evaluation document outlining the narrow regimes where complex deep architectures provide legitimate predictive advantages (e.g., high-frequency tick LOB data).

You'll learn

• Methodological audit — formal verification of mathematical proofs, data provenance, and empirical assertions in a research paper
• Winsorization / Truncation — statistical manipulation of extreme outliers that can inadvertently eliminate real fat-tail risks
• Clustered standard errors — econometric adjustment accounting for correlation across time and across firms

How to work through it

1. 1Select a target peer-reviewed quantitative paper.
2. 2Step through all mathematical proofs and derivations in the paper, noting any unstated boundary conditions.
3. 3Evaluate data cleaning and construction: check for survivorship bias, point-in-time adherence, and winzorization choices.
4. 4Examine econometric tables: verify whether standard errors are cluster-adjusted, check t-statistics against Harvey's thresholds, and inspect R-squared plausibility.
5. 5Compile findings into a structured audit memo.

You'll learn

• Investment Committee Due Diligence Memo — formal institutional report evaluating strategy credibility, capacity, risks, and net return expectations
• Quant Quake (August 2007) — historic multi-day fire-sale unwind of equity market neutral statistical arbitrage models
• Adversarial model testing — intentionally subjecting a quantitative model to extreme regime shifts and friction stresses to uncover hidden failure vectors

How to work through it

1. 1Extract the core strategy hypothesis, target asset universe, leverage profile, and execution horizon.
2. 2Perform stress tests for transaction costs, market impact, short borrow constraints, and factor crowding.
3. 3Apply the Deflated Sharpe Ratio calculation to discount for backtest trial iterations.
4. 4Evaluate behavior during historical liquidity crises (e.g., 2007 Quant Quake, 2008 GFC, 2020 Covid liquidity shock).
5. 5Formulate a final defensible investment committee memo.

You'll learn

• Portfolio Insurance — dynamic hedging strategy whose collective automated selling exacerbated the 1987 stock market crash
• LTCM (Long-Term Capital Management) — hedge fund collapse demonstrating the catastrophic failure of correlation and liquidity assumptions under extreme leverage
• Research manifesto — definitive set of operational and epistemological principles guiding reliable quantitative investigation

How to work through it

1. 1Draft the core sections: 1. Continuous Math & Stochastic Limits; 2. Microstructure & Execution Realities; 3. Factor & Alpha Identification; 4. Econometric Integrity & Overfitting Prevention.
2. 2Include concrete examples of historical failures (LTCM, 1987 Portfolio Insurance, 2007 Quant Quake).
3. 3Present the manifesto in a live discussion or written exchange with a quantitative practitioner or study partner.
4. 4Incorporate critical feedback into the final version of the document.