Building CetaResearch. Data analysis for policy, research, and startups.
backtests
We tested the Oversold Quality strategy (Piotroski >= 7 + RSI < 30) on 16 stock exchanges. It underperforms SPY everywhere, but the gap ranges from -0.47% in China to -12.19% in Thailand. Cash rates explain most of the variation.
backtests
We combined Piotroski quality scores with RSI oversold signals and tested it on 25 years of US market data. The strategy underperforms SPY overall but shows defensive characteristics in bear markets.
backtests
Oversold Quality on German XETRA stocks: 5.76% CAGR, Sharpe 0.193, and only 59.6% down-capture vs SPY. The best risk-adjusted result in our 16-exchange series, with a significant caveat: 35% of quarters the portfolio held cash.
backtests
The Oversold Quality strategy on Chinese A-shares (SHZ+SHH) delivered 7.54% CAGR vs 8.01% for SPY over 25 years. The standout stat: 12.5% down-capture ratio during SPY drawdowns.
value investing
We ran the same P/B screen across 18 stock exchanges and 25 years of data. Four beat SPY. Fourteen didn't. Here's where the value premium survived and where it didn't.
value investing
We backtested a P/B < 1.5 + ROE > 8% screen on US stocks from 2000 to 2025. 7.50% CAGR vs 7.64% for the S&P 500. The Nobel Prize-documented premium has been arbitraged away. The down capture story (79.3%) is still real.
value investing
We backtested a P/B < 1.5 + ROE > 8% screen on Stockholm-listed stocks from 2000 to 2025. 11.93% CAGR, Sharpe ratio of 0.447, and a down capture of only 51.56%. Best risk-adjusted result of 18 exchanges we tested. Returns in SEK.
value investing
We ran Graham's net-net strategy on 7 global markets over 24 years. Japan matched the S&P 500 with lower drawdown. Hong Kong lost 82% peak-to-trough. The US underperformed by 3.8% per year. Here's the full cross-market comparison with year-by-year data.
value investing
We backtested Graham's net-net strategy on all US exchanges (NYSE, NASDAQ, AMEX) from 2001 to 2024. The result is 5.02% CAGR vs 8.84% for the S&P 500. The premium that Oppenheimer documented in 1986 has largely disappeared. Here's what the data shows.
backtests
We tested EV/EBITDA below 10x on 18 global exchanges with local benchmarks. 14 of 18 beat their local index. The value premium is broader than SPY comparison suggested.
earnings
Taiwan shows delayed post-announcement drift that builds to +1.79% by day 21. Streak 3 reaches +2.60% T+21 and +2.75% T+63. 3,721 events from 2012-2025. The muted day-1 reaction followed by strong drift suggests gradual information incorporation.
earnings
Japan shows the highest T+1 abnormal return of any exchange globally. 5,461 events on JPX from 2009-2025. Streak 3 reaches +2.35% at T+63. The pattern is consistent with Japan's concentrated analyst coverage and slower information diffusion.