Brier: 0.246 across 87 signals (conviction-score based).
Lower is better; 0.250 = random guessing. Scored since Feb 14, 2026.
| Tier | N | Win% | Avg R |
|---|---|---|---|
| High Conviction | 21 | 66.7% | +0.85R |
| Conditional | 43 | 61.0% | +0.93R |
| Low | 106 | 34.8% | +0.13R |
* Low sample (N < 5)
Compare conviction-tier portfolios side by side on one timeline.
| Model | N↓ | Win%↕ | Day↕ | Swing↕ | Position↕ |
|---|---|---|---|---|---|
| DeepSeek-R1 | 139 | 50.4% | 53.3%(8W-7L) | 50.5%(53W-52L) | 47.4%(9W-10L) |
| Gemini-3-Pro | 106 | 50.9% | 60.0%(9W-6L) | 48.7%(38W-40L) | 53.8%(7W-6L) |
| Claude-Opus-4.5 | 105 | 55.2% | 65.0%(13W-7L) | 52.2%(36W-33L) | 56.3%(9W-7L) |
| Claude-Sonnet-4.5 | 95 | 53.7% | 66.7%(6W-3L) | 50.6%(40W-39L) | 71.4%(5W-2L) |
| GPT-5.2 | 85 | 44.7% | 42.9%(6W-8L) | 43.9%(25W-32L) | 50.0%(7W-7L) |
| GPT-4o | 40 | 27.5% | 16.7%(2W-10L) | 27.8%(5W-13L) | 40.0%(4W-6L) |
| GROK-4 | 40 | 40.0% | 36.4%(4W-7L) | 43.8%(7W-9L) | 38.5%(5W-8L) |
| Gemini-3-Flash | 39 | 28.2% | 15.4%(2W-11L) | 29.4%(5W-12L) | 44.4%(4W-5L) |
| Qwen3-235B | 36 | 50.0% | — | 50.0%(18W-18L) | — |
| Claude-Haiku-4.5 | 25 | 24.0% | 37.5%(3W-5L) | 22.2%(2W-7L) | 12.5%(1W-7L) |
| GPT-5.4 | 15 | 40.0% | — | 40.0%(6W-9L) | — |
| Kimi-K2-Thinking | 11 | 63.6% | — | 63.6%(7W-4L) | — |
Disclaimer: These results represent hypothetical signal performance based on entry and exit prices, not actual trades. Past performance does not guarantee future results. Results do not account for slippage, commissions, or other trading costs. This is not financial advice.