Building CetaResearch. Data analysis for policy, research, and startups.
pairs trading
We tested 2,579 highly correlated US stock pairs for cointegration. Only 516 passed (20.0%). Utilities led at 31.4%. Here's the Engle-Granger method, sector breakdown, and the counterintuitive finding: lower correlation predicts higher pass rate.
small cap
We backtested small-cap value on Indian exchanges (BSE + NSE) from 2000 to 2025. 22.48% CAGR, beating the Sensex by 10.42% per year. The highest absolute return of any exchange in our 14-market comparison.
pairs-trading
We ran a multi-pair portfolio construction backtest on US stocks from 2005 to 2024. Adding more pairs made returns worse, not better. Transaction costs erode marginal convergence gains, and the best configuration still lost 0.43% annually.
momentum investing
We ran pure 12-month price momentum (no quality filters) on US stocks from 2000 to 2025. The strategy returned 7.15% annually vs 7.77% for SPY, with 159% down capture and a -66% max drawdown. Small CAGR gap, much worse risk profile.
China stocks
We backtested sector-relative EV/EBITDA on 25 years of China A-share data (SHZ+SHH). 4.11% CAGR vs 3.54% SSE Composite, +0.57% excess. But Sharpe 0.039 and a -64.72% max drawdown tell the real story: the signal barely beats a market that barely returns anything. Discounts follow policy, not value.
value investing
A quality-filtered P/S value screen on Stockholm stocks returned 10.96% annually vs 3.17% for the OMX Stockholm 30 index, the strongest excess return vs a local benchmark of any market tested. Sharpe ratio 0.452, the best globally.
pairs trading
We ran the same pairs trading strategy on 12 stock exchanges from 2005 to 2024. Every single market generated near-zero or negative nominal returns. The academic consensus (Do & Faff 2010) that pairs trading profitability declined post-2002 holds globally, not just in the US.
pairs-trading
After correcting for same-bar entry bias and using local benchmarks, multi-pair pairs trading fails to beat any local equity market across all 11 exchanges tested. Zero positive Sharpe ratios. The strategy offers genuine crisis protection at the cost of small annual losses, not alpha generation.
backtests
India's result is worse than it looks: 4.45% CAGR over 24 years on NSE. Compare that to the Sensex, which returned 14.48% annually over the same period. The strategy trailed the local benchmark by -10.03% per year. The Sharpe ratio is -0.08. The strategy
sector rotation
Sweden's sector mean reversion returned 10.79% annually over 26 years and beat OMX30 in 19 of 26 calendar years (73.1%) and 67.31% of quarters, the most consistent record of 13 exchanges. The tradeoff: a max drawdown of -62.27% and a beta above 1.
backtests
We ran the Graham Number screen on XETRA for 25 years. 8.21% CAGR vs 4.45% DAX, a +3.77% annual excess. Beta 1.00 vs the DAX, up capture 125.4%, down capture 79.8%, and alpha +3.77%. A genuine value edge over the local benchmark.
GARP
GARP on XETRA stocks from 2000 to 2025. 9.52% CAGR vs the DAX's 5.12%. 4.40% annual excess, 65% down capture, Sharpe 2.5x the index. Germany is one of the few markets where GARP generates real alpha vs its local benchmark.