| Model | Directional Accuracy | Mean Absolute Error | Prediction Bias | Hit Rate (±5%) | N |
|---|---|---|---|---|---|
| Claude | 55.6% | 30.3% | +14.3% | 18% | 735 |
| Grok | 53.6% | 34.9% | +12.9% | 9.3% | 739 |
| GPT | 52.4% | 34.5% | +7.1% | 6.5% | 740 |
| DeepSeek | 52.1% | 36.2% | +14.4% | 11.1% | 728 |
| Gemini | 51.1% | 34.2% | +14.5% | 13.6% | 738 |