Trading Backtest on GOOG held-out one-hour window (test)
2,077.68Average ReturnGSPO
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GSPO2026.05 | 2,077.68 | 5,721.89 | 187.62 | 66.86 | -1,128.92 | |
| GRPO2026.05 | 2,069.98 | 5,575.11 | 668.53 | 68.86 | -749.61 | |
| PPO2026.05 | -5.25 | 1,498.13 | 2.02 | 32.73 | -8,364.61 | |
| Q-Learning2026.05 | -63.75 | 2,462.16 | 4.16 | 18.46 | -94,733.55 |