When Markets Aren't Random — and What to Do About It
Test the Efficient Market Hypothesis directly. Run autocorrelation, runs, and variance-ratio tests on real data. Decompose risk through a behavioral lens: which anomalies are real signal, and which are just survivorship?
Quants and portfolio managers curious about EMH violations and edge measurement.
By the end you will…
Test weak-form efficiency on any ticker using three rigorous statistical tests.
Quantify autocorrelation, runs, and variance-ratio deviations.
Decompose realized returns into systematic and behavioral components.
Evaluate fund performance through a behavioral lens — is alpha skill or β masquerading?
Famous 'alpha' fund — does it survive a behavioral audit?
Sharpe
0.78
vs 0.66
Treynor
0.052
vs 0.060
Jensen α
−0.6%
vs —
Info Ratio
0.18
vs —
Apply what you learned
Real-world scenarios that pull together the path. Each links back to the Labs you just used.
Case Study
Is Bitcoin's weekend autocorrelation a tradable signal?
Crypto markets trade 24/7, but volume drops 40-60% on weekends. Pre-2020 papers documented mild positive autocorrelation in BTC weekend returns (ρ ≈ 0.05-0.10). Run the Efficiency Tester on BTC-USD with daily data and see what the 5-year picture looks like now. With venues like Polymarket and Kalshi, weekend prediction-market liquidity has tightened the gap. Result: lag-1 ρ has decayed to ~0.02, no longer statistically significant after transaction costs. Lesson: documented anomalies often disappear once they're known and capital chases them.
The 'low-vol anomaly': real signal or post-hoc rationalization?
Multiple papers show low-volatility stocks outperform high-vol stocks risk-adjusted (a violation of CAPM). Run risk decomposition on USMV (low-vol ETF) vs SPHB (high-beta ETF): USMV's β is ~0.7, SPHB's is ~1.4. Backtest 10 years: USMV returned 9.5% with σ=11%, SPHB returned 12% with σ=22%. Sharpe ratios: 0.50 vs 0.36 — the low-vol anomaly held over this window. But the gap shrunk in the 2021-2023 'meme' regime. Lesson: anomalies persist when behavioral biases (lottery preference, leverage aversion) are durable; they fade when they're mispriced into ETFs.