Seeer AI · Research Data
The definitive reference for which signals work and which are noise. Every entry is backtested with full statistical significance testing — not gut feeling.
9
Signals tested
393
Total events analyzed
25+
Years of data
6 of 9
Statistically significant
3 of 9
Pure noise
Click any column header to sort
| Signal Type | Instrument | Signals | Best Horizon | Win Rate | p-value | Verdict | Report |
|---|---|---|---|---|---|---|---|
| Death Cross + RSI<35 65.2% win rate at 180 days; near-term (30/90d) not significant | Crude OilCL=F | 46 | 180 days | 65% | p=0.003 | Significant | Full report |
| RSI Oversold (<30) 71.8% win rate at 30 days; edge diminishes significantly by 90 days | GoldGC=F | 39 | 30 days | 72% | p=0.0013 | Significant | Full report |
| RSI Oversold (<30) 61.4% win rate at 30d only; edge fades by 60–90d (p>0.10). Best-to-worst spread: +99% to -44% | Natural GasNG=F | 44 | 30 days | 61% | p=0.016 (30d) | Significant | Full report |
| Golden Cross 66.7% win rate and +11% mean at 180d; near-term (30d) is a coin flip with 33% win rate | SilverSI=F | 18 | 180 days | 67% | p=0.031 (180d) | Significant | Full report |
| Technical Breakdown Contrarian signal — breakdowns resolve higher 63% of the time at 90d; mean +4.66% | Crude OilCL=F | 111 | 90 days | 63% | p=0.02–0.04 | Significant | Full report |
| Death Cross Contrarian bullish at 180d — not significant at 90d (p=0.27); mean +5.69% | GoldGC=F | 16 | 180 days | 75% | p=0.03 (180d) | Significant | Full report |
| Death Cross 57.1% win rate at 180d but still -3.1% below unconditional base rate — no edge | CopperHG=F | 21 | 180 days | 57% | p=0.36–0.59 | Noise | Full report |
| RSI Overbought (>70) Coin flip across all horizons — median 90d return is -0.58% despite +0.92% mean (outlier skew) | GoldGC=F | 83 | 90 days | 51% | p=0.29–0.52 | Noise | Full report |
| Golden Cross 73.3% win rate at 180d is the highest in our dataset — but p=0.21 and 90d win rate drops to 46.7% | GoldGC=F | 15 | 180 days | 73% | p=0.21 (180d) | Noise | Full report |
What "Significant" means
p < 0.05 — the signal predicts price direction at a rate unlikely to occur by chance across the full backtest window. Significant signals are still risky to trade; they describe tendencies, not guarantees.
What "Suggestive" means
p between 0.05 and 0.15 — there is a directional tendency but the sample size is too small to clear the 95% confidence threshold. Treat as a starting hypothesis, not a trading signal.
What "Noise" means
p > 0.15 — the signal shows no meaningful predictive power across 25+ years of data. Popular technical signals often fall here. The data wins over conventional wisdom.
Embed this table
Add a live version of this table to your blog or newsletter with one line of HTML:
<iframe src="https://seeerai.com/embed/research-data" width="100%" height="520" frameborder="0" style="border:none;"></iframe>
For custom backtesting on any condition
Ask any question about technical signals or macro events — Fed decisions, ECB rate hikes, supply shocks, tariff cycles — across any commodity or equity. Statistical significance testing included. 10 analyses free, no credit card required.
Start Free Trial →