Before you build a strategy, see how often ideas of the same kind held up on new data. Start from the kinds that held and skip the rest. Each row links to the evidence by asset class. The published base rates draw on 2,825 of the 10,706 out-of-sample tests in the register.
By mechanism
| Mechanism | Tests | Beat benchmark | Held after looking good | Verdict |
|---|---|---|---|---|
| Lead-lag signals | 412 | 26% | 34% | No consistent edge |
| Combined signals | 382 | 30% | 39% | No consistent edge |
| Macro timing | 315 | 41% | 25% | No consistent edge |
| Positioning signals (CFTC and others) | 302 | 30% | 43% | No consistent edge |
| Forced-flow signals | 288 | 24% | 57% | No consistent edge |
| Liquidity and credit signals | 258 | 28% | 30% | No consistent edge |
| Valuation timing | 182 | 42% | 77% | No consistent edge |
| Supply and demand data | 161 | 44% | 44% | No consistent edge |
| Price rules (trend and momentum) | 130 | 36% | 28% | No consistent edge |
| Other signals | 128 | 41% | 54% | No consistent edge |
| Calendar effects | 126 | 16% | 100% | Dead end |
| Behavioural signals | 88 | 43% | 46% | No consistent edge |
| Risk premium harvesting | 26 | 46% | 33% | No consistent edge |
| Attention and news signals | 24 | 21% | 33% | No consistent edge |
By trading frequency
| Frequency | Tests | Beat benchmark | Verdict |
|---|---|---|---|
| Intraday | 347 | 18% | Dead end |
| Monthly | 1790 | 32% | No consistent edge |
| Rare (three or fewer trades per year) | 628 | 40% | No consistent edge |
| Weekly | 33 | 30% | No consistent edge |
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.
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-ALL, as of 2026-11-01.
Questions
What is a base rate for a trading strategy?
The share of comparable ideas that held up out of sample after costs. It is the expectation to start from before trusting your own backtest.
How often do trading ideas hold up?
Across 2,825 out-of-sample tests of 340 trading ideas, 32% beat their benchmark after costs in 2015–2026. Of the tests that looked good before 2015, 39% stayed ahead.