ConvexPi

Curriculum

Six missions in quantitative equity research. Each builds on the last — from the basic overfitting problem through to real data and execution. All missions run in Google Colab. No local installation required.

Start Mission 1Run as a course

6

Missions

8–12 hrs

Total time

Colab notebooks

Format

OOS Sharpe ratio

Assessment

MISSION 1·60–90 min

The overfitting trap

Why in-sample performance is not evidence

Open in Colab

Learning objectives

  1. 1.Explain the difference between in-sample and out-of-sample performance
  2. 2.Construct a cross-sectional signal from synthetic factor data
  3. 3.Measure information coefficient (IC) and its decay
  4. 4.Submit a strategy and interpret your OOS Sharpe score

Key concepts

Information coefficientRolling ICOOS Sharpe ratioOverfitting ratioCross-sectional ranking

Prerequisites

Basic Python and NumPy. No finance background required.

MISSION 2·60–90 min

The limit-order book

Market microstructure and adversarial trading

Open in Colab

Learning objectives

  1. 1.Describe how a limit-order book operates
  2. 2.Implement a simple market-making agent
  3. 3.Measure adverse selection and inventory risk
  4. 4.Compete against other agents in the Arena

Key concepts

Bid-ask spreadAdverse selectionInventory riskMarket impactPnL attribution

Prerequisites

Mission 1 or familiarity with financial returns.

MISSION 3·90–120 min

Alpha discovery

Systematic search under the multiple-testing burden

Open in Colab

Learning objectives

  1. 1.Apply walk-forward validation to avoid look-ahead bias
  2. 2.Control for the multiple-testing problem when scanning features
  3. 3.Measure signal decay and distinguish structural from incidental alpha
  4. 4.Build a composite signal from uncorrelated sub-signals

Key concepts

Walk-forward validationMultiple testing / p-hackingSignal decaySharpe ratio additivityIC correlation

Prerequisites

Mission 1. Basic statistics (t-tests, correlation).

MISSION 4·90–120 min

Strategy library

Replication, combination, and the factor zoo

Open in Colab

Learning objectives

  1. 1.Replicate canonical factor strategies (momentum, value, quality)
  2. 2.Measure pairwise correlation and diversification benefit
  3. 3.Combine strategies using equal weighting and minimum-variance
  4. 4.Diagnose over-fitting vs. genuine factor exposure

Key concepts

Factor zooPortfolio diversificationMinimum-variance weightingFactor correlationOOS replication

Prerequisites

Mission 3. Familiarity with the research library.

MISSION 5·90–120 min

Real data

From synthetic markets to live equity panels

Open in Colab

Learning objectives

  1. 1.Fetch and clean a real equity panel using yfinance
  2. 2.Identify data quality issues (survivorship bias, stale prices)
  3. 3.Adapt synthetic strategies to real market conditions
  4. 4.Compare real vs. synthetic factor behaviour

Key concepts

Survivorship biasLook-ahead biasPrice adjustmentFactor seasonalityUniverse construction

Prerequisites

Missions 1–3. Recommended: review the anomaly tracker.

MISSION 6·2–3 hours

Advanced agents

Reinforcement learning meets market microstructure

Open in Colab

Learning objectives

  1. 1.Implement a basic RL agent for order execution
  2. 2.Measure and minimise market impact
  3. 3.Combine a alpha signal with an execution layer
  4. 4.Evaluate end-to-end strategy performance

Key concepts

Reinforcement learningExecution optimisationTWAP / VWAP benchmarksMarket impact modelsSlippage

Prerequisites

Missions 1–5. Familiarity with basic RL concepts helpful.

Using this as a course

Each mission is designed to stand alone as a 1–2 hour lab session. For a semester course, Missions 1–3 work well as the first half with Missions 4–6 as the research project phase. Classroom cohorts give students a private leaderboard graded on OOS Sharpe — not in-sample performance.

Create a classroom cohortInstructor getting-started guide