A Self-Study Plan for Quantitative Finance (No MFE Required)

A stage-by-stage self-study plan for quantitative finance: what to learn, in what order, and where each stage is actually taught.

A Master of Financial Engineering costs somewhere in the six figures once you count tuition and a year or two of foregone income. For a small set of career paths, that is a rational purchase: some desks and funds filter resumes by credential, and if you want to price exotic derivatives at a bank, the stochastic calculus is not optional. But if your goal is the day-to-day work of a practicing quant researcher, forming a hypothesis about why a price moves, pulling the data, measuring the effect, and trying to kill your own result, none of that requires a degree. It requires a sequence.

Sequence is what most self-study attempts get wrong. People do not fail at quantitative finance because the material is too hard. They fail because they study it in the wrong order: six months of measure theory before ever touching a price series, or the opposite, copy-pasting a backtest before understanding what could make it lie. Here is an order that works, with an honest note on where to learn each stage and what it costs.

Two ground rules before the stages. First, every stage ends with something you produced: a script, a study, a measured result. Reading and watching do not count as completing a stage. Second, the stages are ordered by dependency, not difficulty; skipping ahead to strategies before research methods is the classic self-study failure, because you end up trusting backtests you have not learned to distrust.

Stage 1: Python and data handling, but do not live here

You need enough Python to load data into a DataFrame, filter it, group it, and plot it. That is a smaller amount of Python than most people think. Codecademy and the official Python tutorial cover it free or cheaply, and pandas is best learned against data you care about rather than from documentation.

The trap in stage 1 is staying in it. Tutorial completion feels like progress, so people accumulate certificates in NumPy and matplotlib while never once querying a market. Set a hard exit criterion: the moment you can pull JSON from an API and turn it into a DataFrame, move on. Everything else you will pick up while doing actual research, which is where it sticks.

Stage 2: market mechanics before math

This is the stage almost every syllabus skips, and it is the one that separates people who find edges from people who fit curves. Before you compute anything, you need a working model of what a price is: a live consensus that already contains all public expectations, which only moves on the gap between expectation and reality. From that follows the key question to ask of any strategy idea: what is the mechanism? Who is forced to act, and why hasn't the effect been arbitraged away?

The free opening module of Systematic Trading with Market Data covers exactly this ground, starting with Price as Consensus, and it requires no account and no payment. It takes about ten minutes to work through the first lesson, and it will save you months of testing ideas that never had a reason to work.

Stage 3: research methods, or how backtests lie

Before you test a single strategy, learn the three standard ways a backtest deceives its author: survivorship bias, delisting bias, and look-ahead bias. Each one systematically flatters results, and each one is invisible unless you know to look. An afternoon spent on survivorship bias and look-ahead bias pays for itself the first time you catch your own study committing one.

The complementary skill is reading a result honestly: checking whether the average return hides a fat left tail, whether the profit is concentrated in one regime, and whether the sample is large enough to mean anything. Academic treatments of this exist on Coursera and in the event-study literature, but the fastest route is running event studies yourself and getting graded on the interpretation, not just the code.

Stage 4: go deep on one strategy domain

Breadth is the enemy at this stage. Pick one domain with a clear mechanism, volatility risk premium, scheduled catalysts like earnings, or event-driven classes like dilution, and take it from mechanism to measured result. Depth in one domain teaches you the general research loop: hypothesis, data, measurement, attack. Once you have run that loop end to end a few times, adding a second domain is fast.

This is also where data becomes the binding constraint. Free sources will carry you through stages 1 to 3, but a real event study needs point-in-time data with delistings included, and that is where free feeds quietly fail. We wrote up the tradeoff in free vs paid data for quant students, and if you want project-shaped ideas with the data already lined up, the MFE capstone dataset guide and these event-study project ideas were written for exactly this stage, degree or no degree.

Stage 5: portfolio, risk, and temperament

A strategy that works is not a portfolio. You need to understand position sizing, correlation between sleeves, and above all negative skew: the strategies that win small and often, then lose big and rarely. Most self-taught traders learn this stage from their brokerage statement. It is cheaper to learn it from data first.

This stage is also where temperament stops being a footnote. Every systematic trader eventually sits through a stretch where the strategy loses and the only honest answer to "is it broken?" is a study, not a feeling. Building the habit of answering drawdowns with measurement, before real money makes the question emotional, is a legitimate part of the syllabus.

Stage 6: automation

The final stage is turning research into a daily process: a script that pulls data, ranks candidates, formats a note, and sends it to you on a schedule. This is plumbing, not mathematics, and it is well within reach once stages 1 to 5 are done. Our post on scheduling a daily data pull in Python covers the mechanics.

What the MFE still buys you

Honesty requires saying it plainly: a self-study plan does not replicate an MFE. It will not get you past a resume filter at a bank, it will not teach you stochastic calculus to a level where you can derive pricing models, and it does not come with a recruiting pipeline. If your target is a sell-side quant seat, the credential may be worth the price. If your target is the ability to research, build, and run systematic strategies with your own capital or in a small team, the six stages above cover the working skill set, and they cost a small fraction as much.

A curriculum that compresses the plan

You can assemble all six stages yourself from free and cheap parts, and this post gives you the map. If you would rather have them in one sequence, Systematic Trading with Market Data is roughly 18 hours covering exactly this arc: mechanisms, toolkit, research methods, strategy domains, portfolio, automation. Every lesson runs real Python against real market data in the browser, and the course grades what your code prints rather than what you watched. The first module is free with no account required; the full course is included with Alphanume Pro, along with the data platform the lessons query. Start with the free module and see whether the teaching style fits before deciding anything.