Career track

Quant

Math, code, signals, and the discipline of out-of-sample honesty. This track ends with a strategy you designed, certified across five market regimes it never saw. Pairs with the Markets & Investing track.

01

Mathematical Foundations

The probability and statistics that quant interviews actually test.

Probability: distributions, expectation, tail risk+

Random variables, expected value, variance, and the distributions markets abuse: normal in textbooks, fat-tailed in reality. Why a 5-sigma day happens far more often than the normal curve says it should.

What school skips: That markets are not normally distributed, and most of quant risk management is about what happens in the tails the model ignores.

Learn: MIT OCW: Intro to Probability

Interviews ask

  • You flip a fair coin until you get heads. What is the expected number of flips?
  • Why do markets have fatter tails than a normal distribution predicts?
Statistics: inference, significance, regression+

Sampling, confidence intervals, hypothesis testing, p-values, and linear regression. The machinery for asking "is this pattern real or noise?" which is the entire quant question.

What school skips: Multiple-comparison abuse. Test a thousand strategies and dozens will look significant by luck alone. Most published trading signals are exactly this mistake.

Learn: Think Stats (free book)

Interviews ask

  • You backtest 100 random strategies and 5 have p < 0.05. What did you find?
  • What does R-squared actually tell you, and what does it not?
Time series: autocorrelation, stationarity, regimes+

Why financial data breaks textbook statistics: observations are not independent, distributions shift over time, and yesterday leaks into today. Stationarity, autocorrelation, and why regime changes wreck naive models.

What school skips: That non-stationarity is the defining feature of markets, not an edge case. A model fit to 2013 conditions is a model of 2013, not of markets.

Learn: Forecasting: Principles and Practice (free)

Interviews ask

  • What does it mean for a series to be stationary, and why do we care?
  • How would you detect that a market regime has changed?
02

Code & Data

Manipulate market data without silently lying to yourself.

Python and pandas for market data+

The lingua franca: DataFrames, time indexing, rolling windows, groupby, joins. Compute returns, moving averages, and drawdowns from raw OHLCV without a library doing the thinking for you.

What school skips: Data hygiene. Splits, dividends, missing days, and survivorship bias in your ticker list will corrupt results before your strategy logic ever runs.

Learn: pandas: 10 minutes to pandas · Python for Data Analysis (Wes McKinney)

Interviews ask

  • Compute a 20-day rolling Sharpe from a price series. Walk me through it.
  • What is survivorship bias and how does it inflate backtests?
Lookahead bias: the silent killer+

Any calculation that uses information not available at decision time. Signals computed on close prices but executed at that same close, indicators warmed up on future data, normalizations fit on the whole sample.

What school skips: How easy it is to do by accident. One misaligned index in pandas and your strategy knows tomorrow. If a backtest looks amazing, hunt for the leak before celebrating.

Interviews ask

  • Your backtest shows a 4.0 Sharpe. What do you check first?
  • Give three ways lookahead bias sneaks into a pipeline.
03

Signals & Strategies

Turn a market hypothesis into an executable rule.

What a signal is: hypothesis, rule, edge+

A signal is a falsifiable claim about behavior (momentum persists, extremes mean-revert) compressed into a rule (buy when RSI < 30). Edge is the claim being true often enough after costs.

What school skips: That the hypothesis matters more than the indicator. An RSI rule with no behavioral reason behind it is curve-fitting with extra steps.

Learn: Investopedia: RSI · Investopedia: Momentum investing

Interviews ask

  • Propose a mean-reversion signal. What behavioral claim is it making?
  • Why might momentum work? Why might it stop working?
Build a rule-based signal in the Signal Builder
Position sizing and execution for systems+

Rules need sizing rules: fixed fraction, max positions, stop losses. And execution assumptions: fills, spreads, and why a strategy that trades 400 times a year lives or dies on costs.

What school skips: Turnover as a tax. Academic backtests trade free; real strategies pay the spread every round trip, and high-frequency rules often net out negative after costs.

Interviews ask

  • Your signal trades daily. Costs are 10bps per round trip. How much alpha do you need just to break even?
04

Backtesting Honestly

Stop your own thumb from resting on the scale.

Overfitting: torturing data until it confesses+

Every parameter you tune to history is a bet that history repeats exactly. In-sample vs out-of-sample, walk-forward testing, and why fewer parameters beat more.

What school skips: That iterating on a backtest IS fitting to the sample, even when each iteration feels like an honest improvement. The tenth version of your strategy has seen the test answers nine times.

Learn: Bailey et al: The Probability of Backtest Overfitting (paper)

Interviews ask

  • You improved a strategy through 20 backtest iterations. What is wrong with the final number?
  • Explain walk-forward testing to a non-quant.
Regime robustness: does it survive a different world?+

A strategy fit to a bull market is a bull-market strategy. Real validation runs the rule across regimes it was not designed in: crisis, recovery, high-vol chop, and grinding bears.

What school skips: That the only honest test is data your process never touched. Regimes you cannot pick and cannot see are the cleanest form of out-of-sample that exists.

Interviews ask

  • Your strategy has a 2.1 Sharpe from 2016-2021. Why am I skeptical?
  • Design a validation protocol for a strategy you were handed with no history of its development.
Backtest a signal, then see how it does on a hidden regime
05

Prove It

Certify a strategy across markets it never saw.

Iterate in the lab+

Build signals, backtest them on practice regimes, and study what breaks. The goal is not one hot backtest; it is understanding which of your rules carry real logic.

What school skips: Reps on a real engine. Most students never run a single systematic strategy end to end before claiming quant interest in interviews.

Interviews ask

  • Walk me through a strategy you built. What did you try that failed?
Build and backtest signals free in the lab
The certified signal+

Certification runs your rule across a fixed battery of five hidden regimes: crisis, recovery, volatility, bull, bear. Aggregate return, alpha, Sharpe, and worst-case drawdown land on your public record. Deterministic, so there is no lucky reroll.

What school skips: A systematic record that cannot be cherry-picked. Anyone can screenshot their best backtest; a fixed out-of-sample battery scored server-side is a different kind of claim.

Interviews ask

  • Your certified signal lost money in the crisis regime but beat the battery average overall. Defend it.
Certify a signal across the 5-regime battery
Compete: the Hidden Market Challenge+

Discretionary and systematic minds trade the same hidden market with the same $1M book. Sharpe-ranked, one attempt. Your process against everyone else’s, under identical uncertainty.

What school skips: Ranked evidence. A placement is a line on a resume that a quant interviewer can actually check.

Interviews ask

  • How did you decide position sizes in the challenge without knowing the regime?
Enter the Fall Challenge (Oct 5 - Nov 20)

The map ends in proof.

Work the stages, then put a certified track record behind your resume. Free to start, no card.

Start building yours