Introducing Deep Learning Statistical Arbitrage for after-work quants

Takeaway. Classic pairs/stat-arb hunts mean-reverting residuals after a factor filter; this paper asks whether a deep net can learn those residuals and trading rules—an after-work research template, not a live arb bot.

Deep Learning Statistical Arbitrage (Jorge Guijarro-Ordonez, Markus Pelger, Francesco Zanotti, 2021) is a readable ML take on residual trading.

Fit for learners: Open-PDF classic/working paper on ml-finance — useful for after-work Python quant learners building intuition before coding.

This Hub note is a plain-language intro. Not investment advice. Keep scepticism nearby: Harvey–Liu–Zhu.

The problem

Linear factor filters leave residuals that sometimes mean-revert. Hand-crafted pairs, PCA, and cointegration tests are brittle: they miss nonlinear structure, ignore costs, and overfit quiet regimes.

The idea / core intuition

• Use deep learning to extract a residual / spread process from a large cross-section, instead of only linear factors.
• Map those residuals into a trading policy (when to be long/short the residual book) with realistic frictions in mind.
• Evaluation is about out-of-sample residual quality and costs, not a training-set Sharpe screenshot.
• Capacity of the net is a feature and a leakage risk—walk-forward splits still rule.

Research pattern—not a recommendation to run unfunded pairs books.

Why after-work Python learners care

You already know “spread = y − βx.” This paper is the next sentence: can a net replace β with a flexible filter? Great motivation once pandas and NumPy residuals feel boring.

Tiny evening exercise: take two toy series x, y = randn(200), 0.8*x + 0.1*randn(200); print residual std of y - 0.8*x; write why a neural β that sees the future window would cheat—learning only.

One-evening reading checklist

1. State what “residual” means after a factor filter.
2. Name what the deep model is trying to learn (spread vs policy).
3. List two pitfalls (costs, leakage, regime breaks) for ML stat-arb.
4. Write one linear baseline you would demand before a DNN spread chart.

Download PDF

Download PDF

Landing: arXiv abs

Important disclaimer

Not investment advice. This Hub note is educational only. The paper is academic ML research—not a recommendation to trade statistical-arbitrage signals with real capital.

What this is not

• Not CEF / subsidy framing
• Not Quant or Macro course CTA
• Not a “risk-free arb” product pitch

Next step

Single course CTA: SoloQuant Academy’s Introduction to Python Programming—lists, functions, and files first; residual matrices make far more sense afterward.