Check a signal against random ones before you build on it. Every signal family ran on the same 135 markets as a set of random placebo signals. Placebo baseline: 4.0% significant before 2015, 53% kept direction after.
| Signal family | Signals | Significant before 2015 | Kept direction after |
|---|---|---|---|
| International data | 16 | 4.4% | 55% |
| Cross-border capital flows | 6 | 5.3% | 11% |
| Regulatory and corporate filings | 6 | 6.5% | 31% |
| Expectations and surveys | 11 | 2.8% | 54% |
| Economic expectations | 2 | 3.6% | 49% |
| ETF flows | 17 | 4.2% | 58% |
| Calendar | 3 | 38.6% | 39% |
| Liquidity | 1 | 1.9% | 48% |
| Foreign central banks | 3 | 4.7% | 23% |
| Positioning | 2 | 3.6% | 44% |
| Retail positioning | 3 | 8.0% | 92% |
| Real economy | 3 | 6.0% | 52% |
| Freight | 2 | 2.4% | 14% |
| Semiconductors | 1 | 1.9% | 0% |
| International real economy | 1 | 4.4% | 21% |
| Government finance | 3 | 1.9% | 48% |
| Government and real economy | 1 | 3.4% | 38% |
| Uncertainty and news | 35 | 5.2% | 54% |
Caveats
The 2015–2026 test window has been looked at many times during research, so we label it and do not call it a clean holdout. "Beat benchmark" compares each test with a fixed asset mix or buy-and-hold in the same window; a weak benchmark flatters the result. Many tests use US-listed ETFs. Markets in the matrix are strongly correlated.
Terms used: base rate, out-of-sample test, placebo test, dead end.
Let your AI check how ideas like yours did before it runs a single test. In our pilot, agents that checked first spent 1% of their backtests on known dead ends instead of 15%, and every backtest not wasted is time and tokens you keep. Cite this entry as ER-SIG, as of 2026-11-01.
Questions
Do macro signals predict market returns?
On our evidence, mostly no: they hold about as often as random placebo signals after 2015.