Quant Bible (MIT Sloan Business Club)

A 51-page LaTeX-set study guide written by students of the mit-sloan-business-club at massachusetts-institute-of-technology to prepare undergraduates for quantitative-finance interviews. Compiled from PDF metadata as 17 January 2022; the lead author started it “sometime in my junior fall during an interview season for quantitative finance that I found super-challenging,” and states the goal plainly: “to democratize quantitative finance as a career for the SBC community” — against the field’s reputation as an in-group for geniuses, which the guide argues becomes true as a self-fulfilling prophecy.

This is the spoke’s first source on the human and institutional layer of quant trading: who the firms are, what they hire for, and how the field filters entrants. Every prior source here is a codebase or a system.

What’s in it

Seven sections, and the split is itself informative. Roughly half the page count is textbook review, half is interview machinery.

1. Introduction — a 27-firm application list, an MIT course-road (18.600 probability, 18.06 linear algebra, 14.32 econometrics, 18.650 statistics, 6.006/6.046 algorithms, 6.034/6.036 ML; 18.615 stochastic processes flagged as “basically the main field of mathematical finance”), and a reading list.
2. Probability fundamentals — conditional probability and Bayes, expectation and variance, random variables and their distributions, joint distributions, covariance/correlation.
3. Stats fundamentals — LLN, CLT, confidence intervals.
4. Quant research / data science — least squares vs. nearest neighbours, regression, dimensionality reduction (ridge, lasso, elastic-net, forward/backward stepwise), regression brainteasers, and an econometrics section on causal inference: selection bias derived algebraically, and randomization as the fix.
5. Quant research case studies — three open-ended modelling cases recalled from real interviews (Two Sigma’s NY housing prices, QuantCo’s opera house, Two Sigma’s CitiBikes), worked as feature-engineering narratives rather than closed-form answers.
6. Market making — theory and worked interview games; see market-making.
7. Question bank — brainteasers organized by firm across ten shops (Jane Street, Virtu, Optiver, Akuna, Citadel, Hudson River Trading, Two Sigma, Five Rings, SIG), with solutions.

Key takeaways for this wiki

• Market making is stated as a job before it is stated as a strategy. The guide’s definition is a service obligation — “always be available to buy or sell a particular financial instrument at the prices they quote” — with the bid-ask spread as compensation for the inventory risk that obligation forces you to carry. The corpus previously had “market making” only as one entry in a strategy taxonomy algorithmic-trading-wikipedia.
• The three inputs to a quote — theoretical value, last traded price, current position — and the rule that a flat book quotes symmetrically while an inventoried one skews to shed risk. This is the first source here that describes how a price is formed rather than how an order is placed.
• The interview is the field’s filter, and it is a market-making simulator. The pricing games (Red Sox wins, the population of Tanzania, the number of windows in SIG’s building) test theo-value generation, confidence-interval sizing under ignorance, position tracking, and inference from a counterparty’s trades — see quant-interview.
• The firms and their roles. Prop shops, quant hedge funds, and bank desks, split across quantitative trading / quantitative research / SWE tracks: quant-trading-firm.
• Regression, not machine learning, is the research idiom. The data-science section is Hastie-flavoured (Elements of Statistical Learning is named “essential”), and the case studies are solved with z-scores, F-tests on groups of correlated variables, and regularized regression, not with the RL and boosting the spoke’s software sources reach for (reinforcement-learning-trading, supervised-learning-trading).
• Causal inference is treated as core (§4.6), which no other source here touches: the health-insurance worked example decomposes an observed difference into average causal effect plus selection bias, then kills the bias term by randomization. Markets don’t permit randomized assignment, and the guide doesn’t say so — but that unmentioned gap is exactly why the corpus’s ML sources fight lookahead bias and data scarcity instead.

Provenance, tier and limits

• T3. Student-authored and self-published, with no editorial or peer review, and its most distinctive content — the per-firm question bank — is recalled interview material, unverifiable by construction and lightly attributed (“by Evan and Brian”). The theory sections are second-hand restatements of standard textbooks (Hastie et al., Angrist-style econometrics) rather than original work; the underlying material is T1-grade, the transmission is not.
• Stale by four years (2022 → 2026). The application list, the role names, and the per-firm question banks are the parts that rot fastest. The market-making theory and the probability/stats material do not date.
• Written to sell the career, not to assess it: the introduction argues quant finance is intellectually exciting and “a highly ethical way of working in the finance industry” because the pay funds philanthropy. Treat the framing as advocacy; the mechanics are the durable part.
• The guide is held in raw/quant-bible-mit-sloan.pdf — the Google Drive link it arrived by is not a durable address.

Cross-spoke context

• ../machine-learning-wiki — §4’s least-squares-vs-kNN, bias/variance, and dimensionality-reduction material is that spoke’s subject (classical ML as an engineering discipline), taught here as a hiring prerequisite. This source was routed by its dominant substance (the trading field and its market-making core), not split.
• ../psychology-wiki — the reading list opens with Kahneman’s Thinking, Fast and Slow, recommended as “relevant to trader thinking styles.” The System 1/2 vocabulary is borrowed here as an application; the psychology itself lives there with daniel-kahneman.
• ../optimization-algorithms-wiki — the regularization/subset-selection menu (ridge, lasso, elastic-net, stepwise) is model fitting, not metaheuristic search; the standing homonym warning applies.

Related

market-making · quant-interview · quant-trading-firm · mit-sloan-business-club · massachusetts-institute-of-technology · algorithmic-trading · synthesis