A student in week one
Start from zero and reach a defensible submission.
The method behind the workspace, taught properly.
31 articles across six parts, a 77-term glossary, and four reading paths. Every article teaches one thing with worked arithmetic, names where it breaks, and links to the page in the workspace that implements it.
01Reading paths
The sceptical reviewer
Everything a result has to survive before it is believed.
The allocator
From a spreadsheet of weights to a stress-tested, explained allocation.
The lecturer
A term plan and the reading that supports it.
02Start here
Validation
Purging and embargoing explained visually
Why the gap between train and test has to be at least as wide as the label horizon, and what purging and embargoing each remove.
Validation
Beating the naive baseline (and what it means when you don't)
Why every forecast has to be measured against last-value-carried-forward first, and how to read a skill score that comes out negative.
Models
Why more models is not better
The arithmetic of multiple testing: why trying more estimators, more features and more thresholds makes the best result look better and mean less.
Practice
Reproducible research: snapshots, hashes and audit trails
How a workspace can prove a result was produced the way it claims, and what a content hash and an append-only ledger actually guarantee.
03Six parts
Part 1
Foundations
What quantitative research is, the loop it runs in, and how a result fools you.
- What quantitative research actually is
- The research loop, end to end
- Signal versus noise: what a result has to survive
- Reading a backtest without fooling yourself
- Common failure modes in retail quant research
Data and features
Preparing financial time series, building features without leakage, choosing a benchmark.
- Preparing financial time series
- Returns, log returns and stationarity
- Feature engineering for financial data
- Look-ahead bias and data leakage: the ten ways it happens
- Benchmarks and why your comparison set matters
Models
Model families and when to use them, linear against trees, hyperparameters, forecasting.
- Classification, regression, clustering, dimensionality reduction, anomaly detection
- Choosing between linear and tree-based models
- Why more models is not better
- Hyperparameters that actually matter
- Forecasting versus prediction: what you can and can't claim
Validation
Chronological splits, purging and embargoing, walk-forward folds, the naive baseline, overfitting.
- Why random train/test splits break on time series
- Purging and embargoing explained visually
- Walk-forward validation, fold by fold
- ROC AUC, worst-fold performance, and reporting honestly
- Beating the naive baseline (and what it means when you don't)
- Overfitting: detection, not prevention
Portfolio and risk
Optimisation from first principles, risk metrics that matter, stress testing, explainability.
- Portfolio optimisation from first principles
- Risk and return metrics that matter
- Scenario and stress testing your allocation
- Explainability: knowing why a model said what it said
- Drawdown, volatility and the limits of historical risk
Practice
Reproducibility, checkable reports, re-runnable workflows, presenting work, teaching a term.
- Reproducible research: snapshots, hashes and audit trails
- Writing a research report someone else can check
- Building a research workflow you can re-run
- From coursework to portfolio: presenting quant work
- A term plan for teaching quant research with a workspace
04Glossary
77 terms, each with its formula, its assumption and where it breaks.
A definition without its failure mode is incomplete. The glossary is anchor-linked, searchable, and every term hovered in an article opens here.
- Sharpe ratioAnnualised excess return over annualised volatility, the standard reward-per-unit-of-risk statistic for a strategy or portfolio.
- PurgeRemoving training rows whose label window overlaps a test block, so a model never trains on information the test period's outcome depends on.
- EmbargoRemoving the rows immediately after a test block, where serial correlation would otherwise let information leak backward into training.
- Deflated Sharpe ratioThe Sharpe ratio adjusted for the number of configurations tried and for non-normal returns, so a result picked from many looks less impressive.
- Walk-forward validationA sequence of out-of-sample test periods with refitting at each origin, so every reported result is on data the model never trained on.
- Naive baselineThe simplest rule that requires no model at all — predict no change, predict the trend continues, predict the historical average — that any predictive claim must beat.
- Look-ahead biasUsing information in a calculation that would not actually have been available at the time the decision it informs was made.
- Worst foldThe single lowest-performing fold in a walk-forward or cross-validated evaluation, the number that should headline a validated result, not the average.
Learn it here. Run it there.
Every concept in QuantLab is implemented in the workspace with the same names. Beta access is free and opens by cohort.